You are on page 1of 53

0022-3530/92 $3.

00

The Geology of the Great Dyke, Zimbabwe:


Crystallization, Layering, and Cumulate Formation in
the PI Pyroxenite of Cyclic Unit 1 of the Darwendale
Subchamber

by ALLAN H. WILSON

Downloaded from http://petrology.oxfordjournals.org/ at Aston University on January 30, 2014


Department of Geology, University of Natal, P.O. Box 375, Pietermaritzburg,
South Africa 3200
(Received 5 April 1990; revised typescript accepted 3 October 1991)

ABSTRACT
The PI layer of the Great Dyke is an ~200 m thick pyroxenite succession in Cyclic Unit 1 and, as
the topmost lithology of the Ultramafic Sequence, represents the transition from ultramafic to mafic
rocks. Of critical importance to this part of the stratigraphy is the strong lateral environmental change
from axis to margin as a result of the flared structure of the Great Dyke. During the formation of the
PI layer the axial zone was underlain by a great thickness of hot ultramafic cumulates whereas the
same layer in the marginal zone progressively offiaps the lower ultramafic layers and is in close
proximity to the underlying wall/floor rocks. Heat loss through the floor was therefore much greater
in the marginal zone than in the axis.
Major lateral variations are observed, with all lithological units and layers thinning towards the
margins of the subchamber together with a progressive change in the form of the cumulates.
Discordant relationships towards the margin between layer types (modal, cryptic, and form) are a
feature of the PI unit which has also been recognized in other parts of the Great Dyke (Prendergast,
1991). Pyroxene compositions show significant variations within an overall fractionation trend and
decoupling occurs between major and minor element components of bronzite, suggesting strong
compositional heterogeneity of the magma. This type of cryptic layering has not previously been
described and is informally called 'cryptorhythmic' layering.
Pyroxene compositional variation is related to reaction and modification by trapped intercumulus
liquid, and few minerals preserve liquidus compositions. A similar situation must exist for most
layered intrusions. The strong dependence of pyroxene compositions on incompatible element
content in the whole-rock shows that the original liquidus compositions were modified by postcumu-
lus overgrowth and reaction with the trapped intercumulus liquid. Well-constrained data arrays
indicate that most cumulates in the PI layer behaved as a closed system with little or no migration of
intercumulus liquid. Liquidus compositions can therefore be deduced and the residual porosity and
degree of postcumulus formation were modelled using a computer program. Residual porosity is
shown to be between 1 and 13% (by mass). Rocks in the marginal facies have a relatively large
proportion of discrete postcumulus phases but instead of representing crystallization of trapped liquid
these are shown to be mainly heteradcumulus phases, i.e., interstitial minerals that have grown largely
by adcumulus processes in equilibrium with the main body of magma. The heteradcumulus
component can be as high as 27%. These phases occur as oikocrysts which give rise to a well-
developed nodular pyroxenite (the 'potato' reef). The formation of the nodules caused local redistribu-
tion of primary sulphide liquid.
The liquid layers which gave rise to cumulates in the marginal facies are shown to be enriched in
iron and incompatible elements compared with the axial zone, indicating that the PI pyroxenite layer
formed by crystallization of a magma which was either compositionally stratified or exhibited a strong
lateral compositional gradient.
[Journal of Petrology, Vol. 33, Part 3, pp. 611-663, 1992] © Oxford University Press 1992
612 ALLAN H. WILSON

INTRODUCTION
This paper describes a 200 m thick succession of bronzitites and websterites which
constitutes the critical transition in the stratigraphy from ultramafic to mafic rocks in the
Great Dyke of Zimbabwe and is referred to as the PI pyroxenite layer. The section
comprises one of the most complex packages of rock types in the Great Dyke and it is also
of major economic importance as it contains several sulphide zones enriched in platinum
(Prendergast and Wilson, 1989; Wilson and Prendergast, 1989).
The Great Dyke is a highly elongate intrusion of mafic and ultramafic rocks some 500 km
long and between 2 and 10 km wide with a feeder dyke at depth. The intrusion comprises
several magma subchambers (Wilson & Prendergast, 1989), and the present study was carried
out in the largest of these, called the Darwendale Subchamber. Cyclic layering of dunite or

Downloaded from http://petrology.oxfordjournals.org/ at Aston University on January 30, 2014


harzburgite grading upwards into bronzitite characterizes the upper part of the ultramafic
sequence. The PI pyroxenite is associated with the stratigraphically highest cyclic unit.
In transverse cross-section the chamber is funnel-shaped and postulated sill-like exten-
sions of the uppermost layered sequence (Podmore & Wilson, 1987) are now entirely
eroded. The shape of the Great Dyke magma chamber has considerable significance for heat
loss and crystallization. The elongate form provides a large surface area at the inclined
wall/floor of the chamber and heat flow would have been significant compared with that
from the roof. Furthermore, the attitude of the layering within the chamber is such that at
higher levels the layers of rock offlap progressively onto the floor/wall of the chamber.
Therefore a rock layer in the marginal environment is in close contact with the floor,
whereas in the axis this same layer overlies a great thickness of ultramafic cumulates. These
two contrasting environments within the PI layer are referred to as the axial and marginal
facies and there is a progressive gradation from one to the other. The uppermost section of
the Ultramafic Sequence in the Darwendale Subchamber is well exposed along a strike
length of > 60 km and provides an ideal opportunity to study lateral variation as the
plunging attitude along the longitudinal axis results in the same layer being exposed both at
the central axis and at the margin (Fig. 1). A comprehensive drilling programme of 183
boreholes was carried out between 1968 and 1972 to delineate the platinum group element
bearing sulphide zones. Study of these boreholes and field exposures permits both vertical
and lateral changes in the PI pyroxenite of the Ultramafic Sequence to be investigated in
detail. Mineral compositions have been determined in >500 samples. This study has
provided insight into the structure of the magma body, and its subsequent influence on the
formation of macro-, medium-, and small-scale layering, as well as having implications for
the genesis of the Pt-enriched sulphide zones within the pyroxenite layer. Lateral variations
across the longitudinal axis and asymmetry of primary features are considered to be related
to the fundamental shape of the magma chamber in this region, which may be attributed to
differences in the nature of the wall-rocks on both sides of the Great Dyke.
Models of primary magmatic controls in layered intrusions can be developed only in
terms of the evidence preserved in the rocks. The approach used here is to consider all those
textural features which would have been influenced by changing conditions of crystal
nucleation and growth (grain-size and mineral proportions, crystal shape and orientation)
in an environment which may also have been physically changing. These are in turn related
to precise determinations of mineral compositions for major and minor components (using
mineral separates) as reflecting the type of cumulus processes and post-liquidus reactions.
THE GEOLOGY OF THE GREAT DYKE, ZIMBABWE 613

Downloaded from http://petrology.oxfordjournals.org/ at Aston University on January 30, 2014


rrn Outline of Great Dyke
"-" Ultramafic Sequence

Gabbronorite
PI pyroxenite
Ultramafic Sequence below PI
Granite/country rock
Faults Sel
Railway a Selous Mine
Road
• Village Borehole/w.
number BH

FIG. 1. General locality map of the Great Dyke and the study area. (A) Location of the Great Dyke in Zimbabwe.
(B) Subdivision of the Great Dyke into chambers and subchambers and delineation of the Mafic and Ultramafic
Sequences. (C) That part of the Darwendale Subchamber studied in detail (this area is outlined on map B).
Boreholes specifically referred to in text and figures are numbered. (Note the position of the northernmost borehole
BH 1 in the axis of the Subchamber.) Field sections (T) and Selous Mine are shown. The areas in heavy outline are
those shown in Figs. 3, 5, and 6. The line X-X' represents the transect line of Fig. 4.
614 ALLAN H. WILSON

MAFIC CUMULATES AND THEIR CLASSIFICATION—TERMINOLOGY


FOR THE PI LAYER
Mafic rocks in layered intrusions are by and large the products of fractional crystalliza-
tion of basic magma contained within large chambers. It was realized more than half a
century ago by Hess (1939) that these rocks do not represent liquid compositions but are the
fractionate which controlled the process of differentiation. Wager et al. (1960) formally
conceptualized these ideas and introduced the theory of cumulates which has, in its various
forms and modifications, remained the cornerstone of the igneous petrology of layered
intrusions.
Morse (1986) summarized the various modifications of the cumulate theory since they
were first proposed by Wager et al. (1960), and provided a detailed appraisal of the thermal

Downloaded from http://petrology.oxfordjournals.org/ at Aston University on January 30, 2014


processes which would result in the two major classes of cumulates—adcumulates and
orthocumulates. Wager et al. (1960) and Wager & Brown (1968) based the cumulate theory
on the concept of gravitational settling of early formed mineral phases, with complete
solidification taking place in a diffusional regime by which liquid trapped within the
cumulates interacted with the overlying body of magma. Morse (1986) drew attention to the
various objections to this overall concept which included the validity of crystal settling
(Bottinga & Weill, 1970; Campbell, 1978; Morse, 1979) and the likelihood that cumulus
crystals may have grown in situ (McBirney & Noyes, 1979).
Irvine (1980) questioned the validity of primary adcumulates and introduced the concept
of infiltration metasomatism by which trapped liquid was physically excluded from the
cumulus pile, which at the same time had a profound influence on the composition of the
cumulus crystals with which it came into contact. Morse (1986) showed, however, that
infiltration metasomatism would lead to the development of orthocumulates rather than
adcumulates. As a consequence of the in situ model for the formation of cumulates, Irvine et
al. (1983) also introduced the concept of lateral growth in magma chambers (rather than
accumulation being restricted to the lowest level of the chamber), and this was elaborated
on by Wilson & Larsen (1985) in their detailed study of the Fongen-Hyllingen layered
intrusion. The importance of compaction in cumulates was also emphasized by Sparks et al.
(1985) and quantified by McKenzie (1984), and Tait et al. (1984) postulated that composi-
tional convection within the cumulus pile may also have been important in the formation of
adcumulates. Annealing processes (Ulmer & Gould, 1982; Cawthorn et al, 1983) could also
reduce the amount of trapped liquid and may even form adcumulates.
In reviewing these various contributions, Morse (1986) stated perceptively that all the
above processes operate when they can, and the question then becomes one of deciphering
the rocks to tell us which is likely to be the dominant process. Campbell (1987) suggested
that mechanisms for the formation of cumulates are either of a primary or secondary nature.
Secondary processes are essentially those that cause expulsion of trapped liquid from the
crystal pile as detailed above, and although such processes may have operated there is
strong evidence to suggest that the porosity of a cumulate is in many cases primary and is
governed by the rates of crystal nucleation and growth. Campbell (1987) concluded that
adcumulates represent equilibrium conditions whereas orthocumulates have not attained
equilibrium.
In spite of the apparent similarities between cumulates of many layered intrusions it is
possible that few formed in precisely the same way. This may not be surprising given the
interplay between the processes of nucleation and growth, the thermal properties of the
magma body as a whole, and the compositional constraints on phase equilibria (Campbell,
1987). Therefore, although there is broad consensus on the meaning of terms relating to
THE GEOLOGY OF THE GREAT DYKE, ZIMBABWE 615

cumulus theory, precise usage by individual workers is in many cases inconsistent. An


attempt is made in this work to quantify the cumulus process that occurred in the PI layer
of the Great Dyke and therefore it is necessary also to define the terms used here. The
definitions given adhere as closely as possible to previous terminologies of Wager et al.
(1960), Jackson (1967, 1971), Morse (1979, 1986), and Irvine (1982), but are presented in a
form specifically to aid description and interpretation of the PI pyroxenite of the Great
Dyke and need not necessarily have universal application.
Parent liquid—that liquid which constitutes the main body of magma giving rise to the
cumulates in question. The composition of the parent liquid does not vary on a small scale
but will continuously change in response to bulk crystallization.
Cumulus crystals—granular crystals which constitute the framework of the rock and
which may be euhedral, subhedral, or anhedral. Cumulus crystals are readily identified

Downloaded from http://petrology.oxfordjournals.org/ at Aston University on January 30, 2014


texturally but no inferences can be drawn regarding their origin.
Intercumulus liquid—the liquid occupying the interstices to the cumulus crystals either as
an open system connected to the overlying body of magma, or as a closed system within the
network of cumulus crystals. In the latter situation the intercumulus liquid is called the
trapped liquid.
Adcumulus component—that part (usually early) of the cumulus crystal which grew in
equilibrium with the parent liquid. In some cases, the entire cumulus crystal network may
have formed by adcumulus growth; in other cases, the 'adcumulus component' of the
crystals may be obscured by overgrowth of lower-temperature compositions or by reaction
with the more evolved trapped liquid.
Orthocumulus component—that part of the cumulus crystal as well as additional late-stage
phases which grew from the liquid trapped between the cumulus grains in a closed system or
evolved infiltrating liquid. Orthocumulus overgrowth may take place on crystals which
initially formed by adcumulus growth.
Porosity—As proposed by Morse (1986), this may be of two types relating to entrapment
of liquid. In the early stages of nucleation and crystal growth the porosity will be very high
(approaching one). Where the liquid is not trapped but is part of an open system linked to
the overlying body of parent magma the porosity would be called the initial porosity. As a
consequence of continued crystal growth the initial porosity will be reduced. When the
system becomes closed the porosity is called the residual porosity and the liquid contained
in the pore spaces is the trapped liquid. In perfect adcumulates the residual porosity is zero.
In the final rock the proportion of trapped liquid would be represented by postcumulus
overgrowth on the original cumulus crystals as well as additional later phases. The residual
porosity cannot be determined by direct observation.
Oikocryst—one or more mineral phases which are texturally interstitial to the cumulus
phases (as defined above) and therefore grew later, but which developed in an open system
by which there was interaction with the parent magma. Oikocryst minerals do not occur as
an evenly distributed mesostasis on a small scale, but as discrete single-phase crystals,
sometimes with well-developed crystal form, which usually contain a smaller proportion of
cumulus crystals than outside the enclosing oikocryst. Where these phases grew in
equilibrium with the parental magma they may be regarded as having adcumulus status. In
such cases, all the intercumulus liquid would have been excluded by the growing network of
cumulus crystals and oikocrysts, and the oikocrysts may be regarded as heteradcumulus
phases as defined by Wager & Brown (1968). In practice, the liquid from which the
oikocrysts grew must be slightly more evolved than the parent magma but this may be a
steady-state situation dominated by replenishment of parent magma in the open system. In
616 ALLAN H. WILSON

this case, a component ol the oikocryst may be regarded as being heteradcumulus. The
initial shape and compositions of the oikocrysts may be obscured by orthocumulus
overgrowth of lower-temperature compositions of the same phase during solidification of
the trapped liquid.
Discrete postcumulus phases—these mineral phases surround and sometimes enclose the
cumulus phases. They include both the late-stage phases which grew from the trapped
liquid as well as heteradcumulus (oikocryst) mineral phases. The proportion of discrete
postcumulus phases may be determined by direct observation.
Postcumulus material—this would include the discrete postcumulus phases as denned
above, as well as the orthocumulus overgrowth on the original cumulus crystals. This term
cannot be quantified without an estimate of the overgrowth.

Downloaded from http://petrology.oxfordjournals.org/ at Aston University on January 30, 2014


The terminology given here deviates from several previous classifications in that it is not the
amount of discrete postcumulus phases that defines the type of cumulate (e.g., Irvine, 1982)
but the relative contributions of the endmember types, i.e., the orthocumulus and adcumu-
lus components (the latter including the heteradcumulus component of the oikocryst
phases). The main problem is determining to what extent cumulus crystals have been over-
grown by more evolved compositions, and how much of the postcumulus material formed
from the trapped liquid. Irvine (1982) noted that estimation of the orthocumulus compon-
ent is facilitated only by the use of indirect methods such as whole-rock concentrations of
incompatible elements.
As the cumulus terminology presented here relates specifically to processes rather than to
observational characteristics, all phases in mafic and ultramafic cumulates may, in principle,
be described in terms of the proportion of the extreme endmembers: orthocumulate and
adcumulate. This is in keeping with the genetic aims of cumulus terminology as originally
proposed by Wager et al. (1960) but without the implication regarding crystal settling,
which is largely irrelevant. This terminology also contradicts their more descriptive use as
suggested by Irvine (1982). The classification is therefore based on the source of the
nutrients to the growing crystals, with the adcumulus component being supplied by the
overlying body of parent magma and the orthocumulus component from the evolving
trapped liquid. Rocks which may be regarded as pure adcumulates are found in a number of
layered intrusions [such as in the Dunite Succession of the Great Dyke (Wilson, 1982;
Wilson & Prendergast, 1987)], but extreme orthccumulates are almost unknown in large
layered intrusions. The most extreme case of an orthocumulate is where a phyric rock
approaches the composition of the liquid from which it formed, and this type may be
restricted to essentially undifferentiated mafic dykes and sills which underwent homogen-
eous nucleation of eutectic phases throughout the body and in which there was no
migration of liquid on a mesoscopic scale. In the development of cumulates from large
layered intrusions it is possible that there may be a rapid transition (both in time and space)
from a dominantly adcumulus process to a dominantly orthocumulus process. It is the
relative contributions of these two processes in the final rock that determines its classifica-
tion, and not the paths or mechanisms by which the processes took place. The estimation of
the relative adcumulus and orthocumulus components in mafic layered intrusions is of
fundamental importance. The term 'mesocumulate', which is commonly used to describe a
rock that has intermediate amounts of postcumulus phases [on the classification of Irvine
(1982), discrete postcumulus minerals would constitute 7-25% of the rock], then has no
meaning in the genetic cumulus terminology.
As noted above, mechanical expulsion of intercumulus liquid (e.g., by compaction) may
assist in the formation of adcumulates but is unlikely to be the only mechanism responsible
THE GEOLOGY OF THE GREAT DYKE, ZIMBABWE 617

in producing this rock type. Even complete expulsion of trapped liquid by compaction
would not produce an adcumulate if overgrowth from the trapped liquid on the original
cumulus grains had already commenced. On the basis of texture alone, it would be
impossible to identify such a rock as not being a perfect adcumulate.

MAGMA CHAMBERS, STRATIGRAPHY, AND STRUCTURE OF


THE GREAT DYKE
Magma chambers
The Great Dyke (2416+ 15 Ma; Hamilton, 1977) (Fig. 1A) comprises two major magma
chambers (North and South Chambers) subdivided into a number of smaller subchambers

Downloaded from http://petrology.oxfordjournals.org/ at Aston University on January 30, 2014


(Prendergast, 1987; Wilson & Prendergast, 1989; Wilson et al., 1989). In the North Chamber
(Fig. IB) the three subchambers are, from north to south: Musengezi (30 km long);
Darwendale (220 km), and Sebakwe (124 km). The South Chamber is divided into the
Selukwe (74 km) and Wedza (100 km) Subchambers. The subchambers are distinguished on
the basis of continuity of layers and style and thickness of cyclic units (Wilson &
Prendergast, 1989). In a previous classification (Worst, 1958, 1960) the Darwendale and
Sebakwe Subchambers were combined to constitute the Hartley Complex. The nomenclat-
ure based on individual complexes has been abandoned (Wilson & Prendergast, 1989). A
possible third magma chamber (the Mavuradona Chamber) has been postulated to exist at
the extreme north end of the Great Dyke (Wilson & Prendergast, 1987).

Stratigraphy and cyclic units


All magma chambers are stratigraphically similar, with a lower Ultramafic Sequence
overlain by an upper Mafic Sequence (Wilson & Wilson, 1981), the latter being preserved as
remnants along the length of the Great Dyke (Fig. IB). The Ultramafic Sequence is broadly
divided (Wilson & Prendergast, 1989) into the lower Dunite Succession and the upper
Bronzitite Succession (shown in Fig. 2A for the Darwendale Subchamber), each containing
well-developed cyclic units. The cyclic units in the Dunite Succession are defined by
repeated occurrences of dunite or harzburgite separated by chromitite layers. Cyclic units in
the Bronzitite Succession have a basal chromitite, or chromite concentration, overlain by
dunite which in turn grades upwards through harzburgite and olivine bronzitite into
bronzitite. This pattern of lithologies is similar to that observed in the Stillwater Complex
(Jackson, 1970) and the Jimberlana Intrusion (Campbell et al., 1970). The nature of the
cyclic units in the Bronzitite Succession differs between the two magma chambers. In the
North Chamber there are six cyclic units, each ~ 2 0 0 m thick, whereas 16 units ~ 8 0 m
thick are developed in the South Chamber (Wilson & Prendergast, 1989).

Structure of the magma chamber and the layered sequence


The structure of the magma chambers and the attitude of the wall-rocks have been
investigated by gravity studies (Weiss, 1940; Podmore, 1970; Podmore & Wilson, 1987). The
wall-rock contacts dip inwards towards the longitudinal axis at between 20 and 35°, but
steepen at depth to become subvertical at the narrow feeder dyke system. A feeder dyke is
postulated (Podmore & Wilson, 1987) to be present along most of the length of the Great
Dyke, and in some parts of the Darwendale Subchamber this feeder may be associated with
deep-seated magma chambers, underlying the floor-rocks. Gravity studies also reveal that
618 ALLAN H. WILSON

CYCLIC
UNIT

u
o
o
500"

Downloaded from http://petrology.oxfordjournals.org/ at Aston University on January 30, 2014


o
o

1000-
o
E
o

1500" ffr^

Gabbronorite
2000 - Olivine gabbro
Websterite
Bronzitite
HTiGabbronorite
Olivine bronzitite
ggjWebsterite L
Bronzitite Granular harzburgite
• Olivine bronzitite
Poikilitic harzburgite
QHarzburgite / dunite
Dunite
hromitlte
Chromitite
(Border group

FIG. 2. Stratigraphy of the Darwcndale Subchamber. (A) Major subdivisions and lithologies of the Ultramafic
Sequence. (B) Detail of Cyclic Unit 1 in the axis of the subchamber, showing subunits and lithologies. (C) Detail of
Cyclic Unit 1 near the west margin of the subchamber and correlation with the axial section. The bronzitites and
websterites of Subunits la and lb comprise the PI pyroxenite.

the only major break in layering continuity along the length of the Great Dyke is at
Lalapanzi (Fig. IB), which marks the separation between the North and South Chambers.
The bronzitites are more resistant to weathering than the dunites (completely serpentin-
ized on surface), and dominate the outcrop in many areas as a series of parallel ridges. The
outcrop pattern indicates the cross-sectional form of the layering to be synformal, with the
same sequence repeated symmetrically about the central longitudinal axis. This results in
exposure of the same layer in the axial and marginal zones because the synformal
arrangement of the layering plunges at a shallow angle (3-5°) from the ends of the
subchamber. In the Darwendale and Sebakwe Subchambers, the layering plunges to the
THE GEOLOGY OF THE GREAT DYKE, ZIMBABWE 619

south and north respectively, resulting in preservation of a remnant of the Mafic Sequence.
Remnants of the Mafic Sequence are also preserved in the Musengezi, Selukwe, and Wedza
Subchambers (Fig. IB). In transverse section, the layering in the Ultramafic Sequence is
considered to be sigmoidal, resulting from the dip of the layering decreasing near the
extreme margin and also close to the axis (Wilson & Prendergast, 1989). This shape
contrasts markedly with traditional semi-circular cross-sections (Worst, 1960), and has
important implications for the relative cooling history of the marginal and axial facies of the
layered sequence.
A critical question in understanding the genesis of the Great Dyke is to what extent the
present synclinal structure and longitudinal plunge of the layering is primary and to what
extent does it reflect postsolidus deformation. Worst (1960) considered the layering in its

Downloaded from http://petrology.oxfordjournals.org/ at Aston University on January 30, 2014


original form to have been subhorizontal with deformation occurring at high temperature
both laterally and longitudinally in a graben structure which now defines the margins of the
Great Dyke. Paleomagnetic studies on the Bushveld Complex (Gough & van Niekerk,
1959) showed that rocks of the Main Zone were deformed by subsidence to their present
position after magnetization. Similar studies on the Great Dyke (McElhinny & Gough,
1963) indicated that no definite answer could be given on whether magnetization preceded
or followed the synclinal formation of the layering. It is entirely possible that the original
attitude of the layering was synclinal but accentuated by later subsidence.

CYCLIC UNIT 1
Stratigraphy and lithologies
Cyclic Unit 1 (Fig. 2B) is the topmost cyclic unit of the Ultramafic Sequence and marks
the important transition from olivine and bronzite cumulates through websterites to
gabbroic rocks of the Mafic Sequence. This cyclic unit is also of major economic importance
as it hosts both mineable chromitite layers and several zones of platinum group element
(PGE) mineralization (Prendergast, 1988; Prendergast & Wilson, 1989).
In the axis of the Darwendale Subchamber, Cyclic Unit 1 is 420 m thick and is subdivided
into six subunits on the basis of chromitite layers, chromite disseminations, and repeated
lithologies (Wilson & Prendergast, 1987, 1989; Prendergast & Wilson, 1989). The four lower
subunits (lc-lf) are each characterized by basal concentrations of chromite overlain by
poikilitic harzburgite that grades into olivine bronzitite. The poikilitic harzburgite is
characterized by reaction and replacement of olivine and poikilitic enclosure by postcumu-
lus orthopyroxene. Granular harzburgites and olivine bronzitites [the distinction between
these two rocks is the relative proportions of olivine and orthopyroxene; see table 1 in
Wilson & Tredoux (1990)], which in turn overlie the poikilitic harzburgite, do not show
olivine reaction.
The overall upward change in the cyclic unit is to dominantly orthopyroxene-bearing
lithologies with a corresponding decrease in modal olivine. The two uppermost subunits (la
and lb) are different from the lower subunits in that pyroxene is the main mineral
constituent. In the axis of the Darwendale Subchamber a narrow (3 m thick) olivine
bronzitite marks the base of Subunit la. This is successively overlain by a 30 m thick
bronzitite layer (Fig. 2B) and a websterite layer up to 37 m thick. Subunit lb comprises
bronzitite with two narrow zones (<1 m thick) of olivine-bearing bronzitite in which relict
olivine grains are entirely enclosed in large cumulus bronzite crystals.
The bronzitite/websterite succession of Subunits la and lb together comprise what is
informally referred to as the PI pyroxenite layer (Fig. 2A and B). The narrow olivine
bronzitite and olivine-bearing bronzitite layers are included within the PI bronzitite.
620 ALLAN H. WILSON

Thickness variation and topology of the PI layer


Data from boreholes (localities shown in Fig. 1C) provide a unique opportunity to
investigate lateral variations in thickness of the layers and cyclic units as well as topological
variations of contacts between the main lithologies. All boreholes were initiated in the
overlying gabbroic rocks and penetrated the websterite and bronzitite layers of Subunit la
and the upper bronzitites of Subunit lb. None intersected the entire PI succession but the
whole sequence is variably exposed in streams, quarry pits, and rail cuttings. Most of the
boreholes were located in the west side of the subchamber in a 13-km strip north of the main
Harare road and parallel to the outcrop limits of the Mafic Sequence (Fig. 1C). In a
northward direction this strip is located progressively closer to the axis. Two boreholes were
drilled in the axis in the northern extremity of the Mafic Sequence. Seven boreholes were

Downloaded from http://petrology.oxfordjournals.org/ at Aston University on January 30, 2014


drilled on the east side of the subchamber, but there are too few in this region to delineate
small-scale variation. Variations in vertical thickness of the websterite layer are shown in
the contour diagram of Fig. 3. Detail is lacking towards the axis and at the east margin, but
complex variations are observed at the west margin, where there is extensive borehole
coverage. The thickness contours are essentially parallel to the geometric axis and show an
overall thinning of the websterite layer away from the axis and towards both the east and
west margins. However, there also exists a broad asymmetry in that the websterite layer at

Thickness of websterite

FIG. 3. Contoured variation of thickness of websterite (metres). Outline of area shown in Fig. 1C. Lack of detail in
the axial environment and at the east margin is due to paucity of borehole cover for these areas. The borehole
intersections are corrected for dip.
THE GEOLOGY OF THE GREAT DYKE, ZIMBABWE 621

the east side is considerably thicker ( > 17 m) than that at the west side ( < 9 m). Irregularities
in the thickness of the websterite layer on the west side appear to be channel-like, and
orientated perpendicular to the margin of the subchamber. Similar irregularities probably
occur on the east side but cannot be resolved because of the relatively poor borehole
coverage.
The thickness changes in the websterite layer are not linearly proportional to the distance
from the axis, and the greatest rate of change occurs at a position intermediate between the
axis and the margin. Measurements of dip and outcrop width of the PI bronzitite indicate a
similar marked change in thickness from 190 m in the axis (at the Manyame River; Fig. IB)
to 142 m at the west margin in the vicinity of Selous Mine (Fig. 1C), representing a change in
thickness of some 25% (Figs. 2B and C). In contrast, the minimum thickness of the

Downloaded from http://petrology.oxfordjournals.org/ at Aston University on January 30, 2014


bronzitite layer on the east side is estimated at 160 m.
Irregularities in the thickness of layers and topological variations were investigated on a
longitudinal transect along a 13-km line parallel to the outcrop of the mafic contact and
220 m to the east of the contact. This section line (shown by line X-X' in Fig. 1C) was chosen
because it represents the highest density of borehole coverage. In the analysis of boreholes
in the transect, several topological highs are indicated with local (over a distance of ~ 1 km)
vertical differences of 20-30 m (Fig. 4B). The thickness of the websterite layer (Fig. 4A)
increases generally from 11 to 21 m, reflecting the progressive change to the thicker axial
facies in the north. Superimposed on the progressive change are local variations in the
thickness (up to 7 m) of the websterite layer, and the variation is such that the layer has a
tendency to be thickest in the troughs (Fig. 4B).

SOUTH NORTH
mO-

5-
10- Pi websterite

15-
20-
Pi bronzitite
25-

200
8 10 II 12 13 km

FIG. 4. Topological variation of the gabbro, PI websterite, and bronzitite units on transect line X-X' shown in
Fig. 1C. (A) Irregular contact of the websterite and bronzitite layers shown relative to the gabbro contact. The
general increase in thickness of the websterite layer from south to north reflects the transect coming closer to the
axis. (B) Topological variation of the Pl-gabbro contact. The websterite layer (solid black) is indicated to be
thickest in those areas where the mafic-ultramafic contact is topologically lowest.
622 ALLAN H. WILSON

The origins of the small-scale topological variations and local changes in the thickness of
the websterite layer are not known but suggest control other than the rate of crystallization
of the layers. An additional important field observation is that small-scale layering (1-2 cm
in thickness), 'cross-bedding', and erosion structures (see section on field relations) in the
gabbroic rocks immediately overlying the websterite layer are most common in the areas
where the mafic-ultramafic contact is highest, and where the websterite layer is thinnest
(Fig. 4B). Explanations for the structures are that (1) the layered sequence reflects (possibly
on a reduced scale) the irregularities of the underlying granite floor/wall rocks, (2) growth
faults were developed in the floor rocks early in the development of the chamber and
movement persisted during the crystallization stage, and (3) these relate to post-emplace-
ment subsidence of the floor. These three possible processes may be interrelated and may all

Downloaded from http://petrology.oxfordjournals.org/ at Aston University on January 30, 2014


have occurred, but the present data set does not permit unambiguous assessment. It is
postulated that the build-up of cumulates would occur in low zones whereas higher-standing
zones would be subject to greater scouring and erosion.

Field relations of the PI pyroxenite


The PI pyroxenite comprises dominantly bronzitite with a thin but important websterite
layer developed at the top. Plagioclase is interstitial together with phlogopite, quartz, and
sulphide minerals. A very coarse-grained mafic pegmatoid with clinopyroxene crystals up to
30 mm in length is commonly developed at the top of the websterite layer and immediately
underlying the gabbroic rocks. The pegmatoid, which is developed at the same stratigraphic
position in all subchambers, is variable in thickness but seldom exceeds 0 5 m.
Bronzitite and websterite of the PI pyroxenite exposed in the axis of the Darwendale
Subchamber north of the Harare railway line (Fig. 1C) feature well-developed, medium-
scale (10-25 m thick), small-scale (1-5 m), and centimetre-scale (1-5 cm) layering. The
medium-scale layering is clearly expressed topographically as a series of low wooded ridges
separated by shallow, sparsely vegetated valleys. Five dominant ridges are observed (Fig. 5),
but within these small-scale layering is also clearly discernible on aerial photographs
(Fig. 6). Small-scale and centimetre-scale layering (Fig. 7) is best developed in the uppermost
20 m of the PI bronzitite on the west side of the axis. The dominant characteristic of the
layering on all scales is that the resistant lithology is slightly more feldspathic and coarser
grained than the more readily weathered bronzitite.
Small-scale layering is also pronounced west of the subchamber axis, where the websteri-
te consists of alternating coarser-grained (grain-size 1-5—2 mm) and finer-grained (grain-size
0-5-1 mm) units 8-12 cm in thickness. The coarser-grained layers are augite rich whereas
bronzite is the dominant mineral in the finer-grained layers. Large megacrysts (up to 3 cm in
diameter) of bronzite occur at the bases of the finer-grained layers within the bronzitite and
websterite, and decrease in abundance upwards in each layer. The small-scale layering
observed in outcrop and boreholes west of the longitudinal axis of the Darwendale
Subchamber dies out towards the west margin and does not occur on the east side of the axis.
This is also seen in the Jimberlana intrusion, where well-developed layering at the centre dies
out towards the margins (Campbell, 1987). An important observation is that where the
small-scale layering is developed in the Great Dyke PI bronzitite immediately underlying
the websterite layer, the thickness of the latter rock type is reduced by an amount
corresponding exactly to that of the small-scale layered bronzitite.
In all areas, the gabbroic rocks overlying the websterite unit are finely layered and exhibit
cross-bedding, slump structures, and contorted bedding. In the vicinity of Selous Mine near
THE GEOLOGY OF THE GREAT DYKE, ZIMBABWE 623

30-28' 29'

17° 46'—

47—

Downloaded from http://petrology.oxfordjournals.org/ at Aston University on January 30, 2014


48'-

\
49'—

50'—

51 —

PI
Bronzitite

52'-
Bronzitite
onzitite "| u mafic sequence
irpentinite J with unit number

Granite contact

53'— , Chromitite as exposed by


trenching with unit number

Faults and major fractures

Railway
„„__ . .•••Appritt. llmm dam
54'-

FIG. 5. Outcrop pattern of rock-types of the Ultramafic Sequence in the vicinity of Darwendale Dam (area outlined
in Fig. 1C). Major cyclic units are indicated (number of cyclic unit given) for serpentinite-bronzitite pairs. Resistant
and easily weathered layers in the PI pyroxenite are delineated and these define the medium-scale layering in the PI
bronzitite. Major chromitite layers are shown by heavy lines and identified according to the nomenclature given in
Fig. 2.

the west margin (Fig. 1C) the websterite layer immediately underlying the gabbroic rocks
displays well-developed erosion structures and truncation of the layering (Fig. 8). Frag-
ments of websterite are also enclosed by the mafic rocks and the coarse pegmatoid layer is
disrupted and absent in many areas.
624 ALLAN H. W I L S O N

Downloaded from http://petrology.oxfordjournals.org/ at Aston University on January 30, 2014

FIG. 6. Aerial photograph showing the outcrop pattern of the medium-scale layering within the PI bronzitite as
depicted in Fig. 5. The resistant and more easily weathered layers within the bronzitite are clearly seen. Fine-scale
layering within the medium-scale layering is indicated by white arrows. Some of the faults which displace chromitite
layers CIc and Cld are shown by dashed lines. This photograph was taken before construction of the Darwendale
Dam (completed in 1976) which now covers part of the PI layer on the east side of the axis in this area.
THE GEOLOGY OF THE GREAT DYKE. ZIMBABWE 625

Downloaded from http://petrology.oxfordjournals.org/ at Aston University on January 30, 2014


FIG 7. Small-scale and centimetre-scale layering in the upper bronzitites west of the axis in the Darwendale Dam
area.

Large (up to 3 cm) clinopyroxene oikocrysts enclosing rounded orthopyroxene crystals


are found in the marginal facies of the PI bronzitite. In some areas, these show strong
alignment and suggest deposition in an unconsolidated crystal pile, possibly by disruption
caused by magma movement. Although they clearly formed later than the cumulus
orthopyroxene, these crystals must have nucleated in close proximity to the interface with
the main body of magma and did not crystallize entirely from the trapped liquid.
Wagner (1914) recognized a distinctive form of the PI pyroxenite characterized by its
coarse nodular appearance in weathered outcrop (Fig. 9) and which he named the 'potato
reef. The nodules vary in size from.2 to 10 cm in diameter and, although present at the same
stratigraphic level in all subchambers of the Great Dyke, they are best developed in the west
marginal zone of the Darwendale Subchamber. The largest nodules occur in a zone 1-2 m
thick at, or immediately below the lower contact with the websterite layer. Early prosp-
ectors of the sulphide zone used this feature as a hanging-wall marker to the platinum-
bearing horizon. It also occurs in the axis of the subchamber and, although considerably
thicker (~ 10 m thick) than at the margin, the nodules are poorly developed and seldom
exceed 5 cm in diameter. The size of the nodules decreases above and below the zone of
maximum development and toward the axis of the subchamber. Their occurrence is related
to the presence of plagioclase oikocrysts and the local distribution of postcumulus phases
which includes sulphide (see next section on petrography).

Petrography of the PI pyroxenites


Bronzitites of the PI layer consist of 70-98% cumulus bronzite, with subordinate to
minor postcumulus augite, plagioclase, quartz, biotite, opaque oxides, and sulphides. Minor
olivine occurs in some parts of the layer. Rocks in the axial zone generally have < 10%
discrete postcumulus phases whereas those on the margins have 10-30% of these phases.
626 ALLAN H. WILSON

West East West East

2m

Morth South West East

Downloaded from http://petrology.oxfordjournals.org/ at Aston University on January 30, 2014


^t ^ _ _ N

c^i D

Sit
Im

West East West East

Im 0-5m
Gabbro- layering and
Mafic pegmatoid Pyroxenite lamination implied
FIG. 8. Erosion features, cross-lamination, and truncation of layering in the gabbro and underlying websterite
layer as observed in Selous Mine. Fragmentation of pyroxenite is shown in (A) and (D), and slumping of the
pyroxenite in (D) and (F). Erosion channels and truncated layering are shown in (A), (C), (D), and (E).

As the proportion of postcumulus minerals decreases, primary growth faces of the cumulus
bronzite become obscured, and they increasingly show mutual interference. Perfect crystal
form is observed in cumulus orthopyroxene where the proportion of plagioclase increases
and it surrounds the cumulus grains.
Throughout the bronzitite unit, augite occurs as postcumulus oikocrysts, commonly with
well-developed crystal faces (Fig. 10A) enclosing bronzite crystals that characteristically are
highly irregular, embayed, and rounded because of reaction. Jackson (1961) described
similar textures from the Stillwater Complex.
Plagioclase is postcumulus (Fig. 10B), occurring as irregular grains 15-60 mm in dia-
meter enclosing several hundreds or thousands of bronzite grains and also some of the
augite oikocrysts. The plagioclase oikocrysts are the controlling influence in the formation of
the nodular pyroxenite ('potato reef). In some cases, a single plagioclase oikocryst forms a
nodule but more commonly a cluster of several oikocrysts comprises a single nodule. The
THE GEOLOGY OF THE GREAT DYKE, ZIMBABWE 627

Downloaded from http://petrology.oxfordjournals.org/ at Aston University on January 30, 2014


FIG. 9 Outcrop of nodular pyroxenite located near the contact of the bronzitite and websterite of the PI layer on
the west marginal zone near Selous Mine. (Note hammer for scale.)

distribution of zoning in the plagioclase oikocrysts forming the nodule is informative. In the
centre of the oikocryst the plagioclase is unzoned, weakly zoned, or irregularly zoned, but is
very strongly zoned towards the margins, where the composition may become albitic and is
associated with microgranophyre. This is particularly so where the plagioclase is in contact
with quartz or phlogopite. The plagioclase is always normally zoned towards the contact
with each enclosed cumulus orthopyroxene grain. In the centre of the oikocrysts this zone
occurs as a very narrow rim but becomes progressively more pronounced towards the
margins of the plagioclase crystals. The zoning pattern indicates that in the central part of
the nodules the plagioclase was close to the primary liquidus composition whereas towards
the margins crystallization was largely controlled by the evolved trapped liquid.
The outermost zone of the nodules contains a slightly higher proportion of cumulus
pyroxene (estimated to be ~ 5 % by volume more than in the interior of the oikocryst) and
also has a higher proportion of late-stage postcumulus phases, including net-textured
sulphide. The sulphide is concentrated around the margins of the nodules, and this feature,
together with the strongly zoned plagioclase and the assemblage of late-stage minerals,
renders the margins of the nodules more susceptible to weathering, giving rise to the
subspherical structures which result in the characteristic nodular appearance of this rock
type. The distribution of sulphide around the margins of the nodules is striking in weathered
zones.
The zoning pattern in the plagioclase and the distribution of the clinopyroxene oikocrysts
indicates that these postcumulus minerals did not form entirely by crystallization of the
trapped intercumulus liquid, but instead largely in an open system linked to the parent
magma body (i.e., as heteradcumulus phases). Final enlargement of the oikocrysts would
have taken place from the trapped liquid and which ultimately also gave rise to the
assemblage of low-temperature minerals.
628 ALLAN H. W I L S O N

Downloaded from http://petrology.oxfordjournals.org/ at Aston University on January 30, 2014


FIG. 10. Photomicrographs showing important textures in the PI layer. Abbreviations: Op—bronzite; Ol—
olivine, Cp—augite; PI—plagioclase; S—sulphide; Bi—phlogopite. (A) Bronzite cumulate from the marginal facies,
showing a large euhedral oikocryst of augite. (B) Bronzite cumulate with postcumulus plagioclase from the more
resistant lithologies in the axial zone. (C) Bronzite adcumulate from the finer-grained layers in the axial zone. (Note
the large bronzite megacryst.) (D) Olivine bronzitite marking the base of Subunit la in the axial zone. (E) Augite
bronzite cumulate from immediately below the mafic contact in the axis. (Note the irregular and embayed margins
of the augite crystals.) (F) Net-textured sulphide forming local mesostasis to cumulus orthopyroxene.

Throughout the PI zone, but particularly in the upper part, minerals generated during
local late-stage fractionation are quartz, alkali feldspar, albite, iron oxides, phlogopite, and
apatite. These minerals are accompanied by concomitant deuteric alteration and replace-
ment of cumulus bronzite by amphibole and talc. Hornblende is an alteration and
replacement product of the augite oikocrysts where these are in contact with late-stage
minerals. The development of these late-stage minerals is overall much greater in the
marginal facies than in the axis, but some pyroxenites in the margins also contain very small
amounts of these minerals, indicating that the distribution is markedly heterogeneous. On
this basis, the orthocumulate components are seen to increase from axis to margin but
cannot be expressed in terms of total proportion of discrete postcumulus phases because
THE GEOLOGY OF THE GREAT DYKE, ZIMBABWE 629

there is clear indication that some proportion of these phases did not form from the trapped
liquid. An increase in the amount of discrete postcumulus phases towards the margins
seems to be a feature of layered intrusions and is seen in the Jimberlana (Campbell, 1987)
and Skaergaard intrusions (Henderson. 1975).
Megacrysts of orthopyroxene at the bases of the finer-grained layers (both in the
bronzitite and websterite units) in the axis of the subchamber are well formed and may be as
much as 1000 times larger (by volume) than the average cumulus orthopyroxene crystals
with which they are associated (Fig. IOC). In rare occurrences, remnant olivine is found in
the centre of these grains.
Olivine bronzitite (Fig. 10D) is developed in borehole section BH1 (Fig. 2) from 60 to
63 m. The cumulus olivine grains are corroded and irregular in form but are not in reaction
relationship with orthopyroxene. Bronzitites containing a few grains of olivine enclosed

Downloaded from http://petrology.oxfordjournals.org/ at Aston University on January 30, 2014


within cumulus pyroxene are present at 80 and 109 m.
Websterite is the uppermost rock type of the Ultramafic Sequence and comprises
cumulus bronzite and augite with postcumulus plagioclase and late-stage minerals. In the
axial environment the coarser-grained websterites contain 30-70% augite whereas the finer-
grained websterites comprise 10-15% augite, with the latter showing a higher degree of
postcumulus development compared with the coarser-grained rocks. The webster-
ite in the marginal environments contains 60-80% augite, which also becomes increasingly
abundant upward in the section. In both the axial and marginal environments the
pyroxenes in the websterite close to the gabbro contact appear highly corroded and
irregular in form (Fig. 10E), and indicate that resorption has occurred.
Sulphide minerals occur throughout the PI unit, but they differ in texture depending on
their modal proportion. Where they occur in very minor proportions ( < 0 1 % ) they are
extremely fine grained (0005-0015 mm in diameter), and are closely associated with late-
stage minerals such as quartz and alkali feldspar. In contrast, where the sulphides are more
abundant (from 1 to 10%) the grains are significantly larger (0-2-3 mm in diameter) and are
intimately associated with cumulus bronzite crystals, commonly forming a local mesostasis
(Fig. 10F). Trapped liquid appears to have been displaced by the sulphide, suggesting its
percolation and entrapment through the crystal pile, particularly at the margins of the
plagioclase oikocrysts which form the nodules (Fig. 11). In these zones, postcumulus silicate

5cm
FIG. 11. Disseminated to net-textured sulphide concentrated around plagioclase oikocrysts which form the
nodules. Drawn from tracing of cut sample from Selous Mine.
630 ALLAN H. WILSON

minerals may be virtually absent but remobilization of sulphide has occurred on a scale
of millimetres so that it is incorporated within altered pyroxene and along cleavages
(Prendergast, 1990; Wilson & Tredoux, 1990).
Small amounts of sulphide are also present within unaltered cumulus pyroxene and in the
unzoned portions of the plagioclase and clinopyroxene oikocrysts, indicating its early
liquation and entrapment. Most sulphide was concentrated into the interstices between the
cumulus grains of pyroxene and the oikocrysts, suggesting small-scale movement and
coalescence of the sulphide liquid in response to the crystallization of silicate phases.

MODAL COMPOSITIONS, GRAIN-SIZE, AND MINERAL FABRIC


Modal variations

Downloaded from http://petrology.oxfordjournals.org/ at Aston University on January 30, 2014


Modal analyses for boreholes BH1 (axis) and BH4 (west margin) show clear differences in
the amounts of all mineral phases. On a relative basis, BH4 (Fig. 12) has less cumulus

O 2 4 6 8 10 12 14 IS 18 0 20 40 60 80100 I0 2 IO» 10« 0.4 0.B 1.2 1.6 20 24


Modal % Modal % S (ppm) Bronzlte mean dimension
(mm)
STRATICRAPHIC COLUMN I Websterlte ^ B Bronzitite w v . ^ Modal-grain size cycles
I- 1 Cryptorhythmic units

FIG. 12. Stratigraphic section for borehole BH4 from the marginal facies, showing rock types and modal mineral
percentages for orthopyroxene (Opx), clinopyroxene (Cpx), plagioclase (Plag), and late-stage phases orthoclase (Or),
apatite (Ap), and quartz (Q). Also shown is whole-rock S content with positions of zones of sulphide enrichment SI,
S2, and S3. Mean crystal-size dimensions are shown for bronzite together with type of crystal fabric. (Note the
position of the boundary between Subunits la and lb.) Modal grain-size cycles are highlighted together with
cryptorhythmic units as defined on the basis of orthopyroxene compositions (see text).
THE GEOLOGY OF THE GREAT DYKE, ZIMBABWE 631

orthopyroxene, and more clinopyroxene, plagioclase, and late-stage minerals than BH1
(Fig. 13). In borehole BH1 the alternating small-scale layering in the websterites and
uppermost bronzitites is also reflected in the modal mineral contents. In BH1 the finer-
grained layers contain less plagioclase and clinopyroxene, and more orthopyroxene than
the coarser-grained varieties in both the bronzitites and websterites.
Distribution of sulfide within the PI layer is of economic importance, particularly as this
relates to PGE mineralization (Prendergast, 1990; Wilson & Tredoux, 1990). Determina-
tions of whole-rock S reflect the distribution of sulphide. In section BH4 (Fig. 12) three
sulphide zones are indicated; the uppermost zone (informally called sulphide zone 1, or SI)
is narrow (2 m thick) and contains up to 4% S (equivalent to ~ 10% sulphide). The
stratigraphically lower sulphide zones (S2 and S3) contain much less sulphide (<0-3% S)
but this is disseminated over wider intervals (S3 is 15 m wide). In the axial borehole BH1

Downloaded from http://petrology.oxfordjournals.org/ at Aston University on January 30, 2014


(Fig. 13), the amount of sulphide is much less than in the margins but again three zones of
sulphide enrichment are recognized. Sulphide zone 2 is poorly developed and the zones are
at a greater distance below the contact with the mafic rocks compared with BH4. The

Orthopyroxene ° Finer-grained
^ | Clinopyroxene • Coarser-grained FABRIC TYPE

Finer Coarser

0 2 0 2 4 6 8 0 20 40 60 80100 O 20 40 60 80100 10* 0.4 0 8 12 16 2.0


Modal % Modal % Modal % S(ppm) Bronzite mean length (mm)
STRATIGRAPHIC ^ ^
COLUMN K 8 3 Websterite |Finely layered bronzilite j Ollvlne bronzltlte I 1 Cryptorhylhmic units
Olivine inclusions In Opx ^•Bronzitile ~ ~ Modal-grain size cycles

FIG. 13. Stratigraphic section for borehole BH1 in the axial zone, showing rock-types and modal mineral
percentages. (Note the contrasting mineral proportions in the finer-grained and coarser-grained layers. These layer
types alternate but are shown as continuous fields for clarity.) Also shown is whole-rock S with positions of zones of
sulphide enrichment SI, S2, and S3. The finer-grained rocks are depleted in S compared with the coarser-grained
rocks. Mean crystal dimensions are given for cumulus bronzite together with an indication of fabric type. (Note the
position of the boundary between Subunits la and lb.) Modal-grain size cycles are highlighted together with
cryptorhythmic units as defined on the basis of orthopyroxene compositions (see text).
632 ALLAN H. WILSON

coarser-grained rock-types of the small-scale and centimetre-scale layers in the upper


bronzitites and websterites of BH1 are significantly enriched in sulphide (Fig. 13) compared
with the finer-grained layers, the average S contents of the two layer types being 2000 and
250 ppm, respectively.

Relationship between discrete postcumulus phases and phosphorus


If the discrete postcumulus phases had developed entirely by crystallization of trapped
liquid, then a broad correlation with incompatible elements such as P would be expected
(Henderson, 1975). This dependence was investigated in four borehole intervals, which were
taken as being representative of the various facies in the PI layer and are shown in Fig. 14.

Downloaded from http://petrology.oxfordjournals.org/ at Aston University on January 30, 2014


The data for borehole BH7 show only a small variation in the amount of discrete
postcumulus phases even for a large range in P content, and extrapolation indicates that
rocks with zero P (i.e., implying that they contained no trapped liquid) would still have a
high proportion of discrete postcumulus phases. In contrast, modal data and P content in
the axial zone show a strong correlation, indicating that in these rocks the discrete
postcumulus phases are largely the result of crystallization of trapped liquid.
Conclusions drawn from these relationships are that for rocks in the marginal facies,
either (1) trapped liquid was removed by secondary processes after crystallization of the
discrete postcumulus phases had taken place, or (2) the discrete postcumulus phases are not
entirely the result of crystallization of trapped liquid but reflect growth of these phases in
equilibrium with the main body of magma, i.e., oikocrysts as heteradcumulus phases. The
first option is highly unlikely, as the proportion of trapped liquid excluded from each sample
would have to be the same to maintain consistency of the data set.

Grain-size variations
Several grain-size cycles for cumulus orthopyroxene are observed in the marginal facies
(BH4; Fig. 12) and in the axis (BH1; Fig. 13). Minor cycles are also present but the sampling
density in the present study is insufficient for these to be characterized in detail. The grain-
size cycles show a strong relationship to the crystal fabrics (see next section). The coarser-
grained rocks in each cycle have poorly developed crystal alignment, whereas the pyroxenes
in the finer-grained rocks have distinct preferred orientations.

30-.
©

J25-
o.
m

o f 20-

I - I 15-
o 10-
"o
8 5-
T
20 80 100 120 140 160 180 200 220 240 260 i!)0 30*0
P (ppm) in whole rock

FIG. 14. Relationship of modal percentages of discrete postcumulus phases with whole-rock P for some
stratigraphic intervals in four borehole sections (positions shown in Fig. 1C). (Note the large intercepts at zero P,
indicating that not all the discrete postcumulus phases are the result of crystallization of trapped liquid.)
THE GEOLOGY OF THE GREAT DYKE, ZIMBABWE 633

Modal proportions of postcumulus plagioclase are related to grain-size, and distinct


cycles are observed in the bronzitites (Figs. 12 and 13). As the grain-size increases upwards in
each of the cycles the amount of plagioclase also increases, indicating that the development
of plagioclase oikocrysts is a primary process depending on the available space, which is
controlled by the network of cumulus orthopyroxene crystals.

Petrofabric analyses and influence of grain-shape


Petrofabric analyses of cumulus orthopyroxene were carried out on nine samples of
bronzitite from the axial region at Darwendale Dam. The most striking feature to emerge
from the study (Fig. 15) is the contrast between samples taken from the two layer types as
identified in the field by their resistance to weathering. Pyroxenes from the more easily

Downloaded from http://petrology.oxfordjournals.org/ at Aston University on January 30, 2014


weathered bronzitites (Fig. 15A) exhibit well-developed lineate lamination fabrics in the
layering plane, with c-axes aligned parallel to the length of the Great Dyke. In this pattern
the distribution of b-axes is dominated by a strong central maximum perpendicular to the
layering plane with a rather poorly developed east-west girdle at right angles to the
alignment direction of the c-axes. This arrangement suggests that, in spite of the preferred
orientation of the [0 10] face in the plane of layering, rotation of the crystals about the c-axis
resulted in thefo-axisgirdle. The distribution of a-axes shows similar but weakly developed
elements, as observed for the b- and c-axes, in that both a peripheral fabric and an east-west
girdle are poorly developed. The b- and c-axes, reflecting maximum flattening and

Zc

FIG. 15. Typical petrofabric diagrams for cumulus bronzite in medium-scale layers, indicating strongly contrasting
features. (A) For more easily weathered layer. (B) For more resistant layer. Axial orientations are represented for a-,
b- and c-crystallographic axes. True north direction (TN), and the trend direction (G) of the Great Dyke and dip
direction (D) are shown. For each sample, 120 crystals were measured; contours are for 0-7, 1-5, 2-9, 5 1 , 6-6, and
8-8% per 1 % area.
634 ALLAN H. WILSON

elongation of the crystals respectively, appear to have exerted the dominant control on the
fabric patterns. By contrast, the petrofabric patterns for more resistant bronzitite layers
(Fig. 15B) reflect weak crystal orientations and planar lamination is poorly developed. There
is a suggestion of weak lineation for c-axes within the plane and fc-axes have a tendency to lie
perpendicular to the plane of layering, but this is much subdued compared with the previous
case.
Aligned elongate crystals are a characteristic feature of crescumulate type textures (Wager
& Brown, 1968) and have been interpreted to be diagnostic of in situ crystallization either
perpendicular to the floor of the magma chamber (Donaldson, 1974; Lofgren & Donaldson,
1975), or parallel to the direction of maximum heat flow in a stagnant environment
(McBirney & Noyes, 1979). In granular cumulates orientation of long axes of cumulus

Downloaded from http://petrology.oxfordjournals.org/ at Aston University on January 30, 2014


olivine in the Rhum intrusion was attributed by Brothers (1964) to alignment by magma
currents, and this explanation may also be the most likely for the Great Dyke.
The contrasting petrofabric patterns of the Great Dyke bronzitites probably reflect the
primary shape differences of the cumulus crystals, with the morphologically anisotropic
crystals adopting preferred orientations in a moving fluid medium. Measurements of crystal
lengths in the crystallographic axis directions (using U-stage techniques; Wilson, 1977)
indicate on average that the dimension ratios (a:b:c) for poorly oriented crystals are
0-79:0-61:1, whereas those for the aligned crystals are 0-63:0-48:1. The aligned pyroxenes
are therefore generally more elongate in the c-axis direction and slightly more flattened in
the fo-axis direction than those showing poor alignment.
Fabric type for the borehole sections BH4 (Fig. 12) and BH1 (Fig. 13) is related to grain-
size. The coarser-grained bronzitites show relatively weakly developed fabric compared with
the finer-grained rocks; again, this may be the influence of primary crystal shape in a flow
regime.

Surface expression of layering type


The surface expression of the medium- and small-scale layering in the Darwendale
Subchamber reflects the combined effects of fabric, grain-size, and the development of
discrete postcumulus phases. The coarser-grained poorly orientated layers are well cemented
by postcumulus phases and are relatively resistant to weathering. In contrast, the finer-
grained rock-types with well-aligned crystals are poorly cemented, and the pronounced
lamination gives rise to fissile jointing and disintegration of the rocks.

PYROXENE COMPOSITIONS
Determination of pyroxene compositions
Accurate determination of pyroxene compositions in coarse-grained mafic and ultramafic
rocks by electron microprobe is limited by lack of homogeneity, which results from effects
such as zoning and subsolidus exsolution. The effect of zoning in pyroxenes from large
layered intrusions is usually minor but exsolution is ubiquitous. Integrated microprobe
analyses combined with broad-beam determinations have been considered to provide
reasonable estimates of composition of Bushveld pyroxenes (Buchanan, 1977, 1979).
However, electron microprobe investigations on Great Dyke pyroxenes using both spot and
broad-beam techniques failed to produce relative precision better than 1-5% for major
components. This problem was even more severe for the minor elements (Ti, Al, Ni, Cr, and
Ca in bronzite), which are strongly redistributed during the exsolution process. Repeated
THE GEOLOGY OF THE GREAT DYKE, ZIMBABWE 635

analyses for these elements failed to produce relative precision better than 10%. To
overcome this problem in the present study and to determine optimum conditions for
analysis, pyroxenes were separated by magnetic and heavy liquid techniques for a range of
carefully determined size intervals. Optimum size-fractions to include exsolution lamellae
but to minimize composite particles were determined to be between 100 and 120 (im. Over
300 pyroxene separates were prepared from field samples and borehole core and analysed in
duplicate by X-ray fluorescence (XRF) techniques with long count times (precision of
analyses is indicated in relevant figures). Ferrous iron was determined on 50 pyroxene
samples by the Wilson method (Whipple, 1974) and linearly calibrated against total iron
(correlation coefficient better than 0-9) to give the following equations:
Augite: %Fe 2 O 3 = 0-228(%FeO)* -0-477

Downloaded from http://petrology.oxfordjournals.org/ at Aston University on January 30, 2014


Bronzite: %Fe 2 O 3 =0-113(%FeO)*-0-452

TABLE 1
Representative pyroxene analyses of the PI layer

Location* Tl BH1 BH1 BH1 BH1 BH1 BH1


Metres^ 224 126-80 82-30 28-96 25-91 25-91 20-51
Rock lype% Bron Bron Bron Bron c. Bronf. Bron m. Bron f.
Pyroxene Opx Opx Opx Opx Opx Opx Opx

SiO 2 5603 55-71 55-43 54-74 55-34 54-98 55-55


A12O3 1-44 1-40 1-26 1-46 1-38 1-37 1-34
Fe 2 O 3 § 051 0-59 0-76 0-82 0-74 0-73 0-76
FeO§ 804 8-71 1006 10-51 9-84 9-76 1000
MnO 018 0-21 0-22 0-24 0-23 0-23 0-24
MgO 31-24 30-91 29-90 29-38 3009 3002 3008
CaO 1-64 1 66 1-71 1-99 1-89 2-29 1-77
Na 2 O — — — — — — —
TiO 2 010 008 014 015 009 010 009
Cr 2 O 3 0-61 0-53 0-49 0-50 0-57 0-59 0-49
NiO 008 008 009 0-07 007 007 007
Total 99-87 99-88 10006 99-86 100-24 10014 100-39

Location* BH1 BH1 BH1 BH4 BH4 BH5 BH5 BH7


Metres^ 20-51 1-52 1-52 74-13 38-16 5-18 5-18 13-95
Rock type\ Bronf. Webst Webst Bron Bron Webst Webst Bron
Pyroxene Cpx Opx Cpx Opx Opx Opx Cpx Opx

SiO 2 53-41 54-91 5313 55-28 54-91 54-33 53-57 54-90


A12O3 1-99 1-30 215 1-39 1-37 1-52 202 1 41
Fe 2 O 3 § 0-63 0-87 083 O82 085 107 102 100
FeO§ 4-31 10-88 500 10-52 1071 12-50 5-64 11-91
MnO 016 0-25 017 O21 024 0-29 018 025
MgO 17-89 2908 17-80 29-51 29-43 27-46 1612 27-80
CaO 2000 211 19-60 1-81 1 87 212 2033 200
Na 2 O 0-30 033 — 025 —
TiO 2 016 013 022 015 013 014 037 018
Cr 2 O 3 0-93 028 051 O47 044 029 056 039
NiO 0-04 O07 004 O07 008 007 004 O09
Total 99-82 99-88 99-78 100-24 10003 99-79 99-64 99-93
636 ALLAN H. WILSON

where FeO* represents total iron expressed in the ferrous st-ate. All pyroxene compositions
were normalized to those equations to allow self-consistent estimates of ferric and ferrous
iron. These equations are considered to be applicable to all pyroxenes in the PI layer of the
Great Dyke.
Representative analyses of pyroxenes are given in Table 1.

TABLE 1 (Continued)
Cations based on 6 oxygens

Location* TI BHl BH1 BHl BHl BHl BHl


Metres^ 224 126-80 82-30 28-96 25-91 25-91 20-51

Downloaded from http://petrology.oxfordjournals.org/ at Aston University on January 30, 2014


Rock typeX Bron Bron Bron Bron c. Bronf. Bron m. Bronf.
Pyroxene Opx Opx Opx Opx Opx Opx Opx

Si 1-962 1-958 1-959 1-946 1-952 1-945 1-956


A1 I V 0038 0042 0041 0054 0-048 0055 0-044
A1 V1 0022 0016 0012 0007 0009 0002 0012
Fe3 + 0014 0016 0020 0022 0020 0019 0020
Fe2 + 0-236 0-256 0-297 0-312 0-291 0-288 0-295
Mn 0005 0006 0007 0007 0007 0007 0007
Mg 1-631 1-619 1-575 1-556 1-582 1-582 1-579
Ca 0062 0063 0065 0076 0071 0086 0067
Na — — — — — — —
Ti 0003 0002 0004 0-004 0002 0-003 0002
Cr 0017 0015 0014 0014 0016 0016 0014
Ni 0002 0002 0003 0002 0002 0002 0002
Total tet. 2000 2000 2-000 2-000 2000 2000 2000
Total oct. 1-992 1-995 1-997 2-000 2000 2005 1-998
mg-number 0-874 0-864 0-841 0-833 0-845 0-846 0-843

Location* BHl BHl BHl BH4 BH4 BH5 BH5 BH7


Metres! 20-51 1-52 1-52 74-13 38-16 5-18 5-18 13-95
Rock typeX Bronf. Webst Webst Bron Bron Webst Webst Bron
Pyroxene Cpx Opx Cpx Opx Opx Opx Cpx Opx

Si 1-947 1-954 1-942 1-955 1-949 1-951 1-955 1-961


A1 I V 0053 0046 0057 0045 0051 0049 0045 0039
A1 V 1 0033 0-009 0034 0013 0007 0-016 0043 0-021
Fe3 + 0017 0023 0034 0022 0023 0029 0028 0027
Fe2 + 0132 0-324 0152 0-311 0-318 0-376 0-174 0-356
Mn 0005 0-008 0-005 0006 0007 0-009 O006 0008
Mg 0-972 1-543 0-970 1-555 1-557 1-470 0-884 1-481
Ca 0-782 0080 0-767 0069 0071 0082 0-802 0077
Na 0-021 — 0023 — — — 0-018 —
Ti 0-004 0003 0006 0004 0-003 0004 0010 0-005
Cr 0027 0008 0-015 0013 0-012 0008 0-010 0011
Ni 0001 0002 0-001 0-002 0002 0002 0016 0-003
Total tet. 2000 2000 2000 2-000 2-000 2-000 2-000 2000
Total oct. 1-993 2000 2-006 1-995 2-000 1-996 2001 1-990
mg-number 0-880 0-827 0-865 0-833 0-830 0-796 0-836 0-806

* BH refers to borehole number (Fig. 1).


t Metres below mafic-ultramafic contact.
X Webst—websterite; Bron—bronzitite; c—coarser-grained layer; f—finer-grained layer, m-
megacryst.
§ See text for determination of FeO and Fe 2 O 3 .
THE GEOLOGY OF THE GREAT DYKE, ZIMBABWE 637

Compositional variation of orthopyroxene in vertical sections


Systematic sampling of the PI layer was carried out to determine vertical and lateral
controls on the major and minor element compositions of cumulus pyroxene. Only three
boreholes penetrated to deeper than the uppermost part of the bronzitite unit, and these
only through the top half of the succession. Nowhere is the entire bronzitite layer exposed
and although sampling is possible in some areas, the bouldery and generally weathered
outcrop does not permit this on a close interval. Continuousfieldsampling was carried out
in the axis of the subchamber at Darwendale Dam (field section Tl and Tla in Fig. 1C) to
investigate the medium-scale layering observed in the bronzitites; samples were also taken
at several localities near the west margin (localities T2-T6, Fig. 1C). Detailed discussion of
the vertical variation of pyroxene compositions will be restricted to thefieldsection Tl near

Downloaded from http://petrology.oxfordjournals.org/ at Aston University on January 30, 2014


the axis, and the boreholes labelled in Fig. 1C. In the Darwendale Dam section (Tl), cyclical
variation in major components [expressed as Mg/(Mg + Fe 2+ ) and referred to as mg-
number] for bronzite (Fig. 16) is identifiable with the medium-scale layering. The rocks of
the poorly exposed, well-orientated, and slightlyfiner-grainedlithologies have consistently
higher mg-number than the associated coarser-grained layers, which are also more resistant
to weathering. Overall, there is a marked decrease in m^-number upwards in the PI layer
from 0-881 at the base to 0-815 at the top. There are no sharp compositional discontinuities,
but relatively smooth cyclical variations, with the mg-number peaking near the middle or
the lower parts of the more easily weathered layers. Minor elements also show marked
cyclical variation superimposed on the overall trends (Fig. 16). Minor elements Ni, Cr, and
Al decrease upwards and reflect the smaller-scale variation of nuj-number. Titanium

I Easily weathered layer Resistant layer

.81 .82 .83 .84.85 .86.87 .88 .07 .08 .09 .09.11 .13 .15
Mg/(Mg+Fe 2 *) %NiO % TiO2

.4 .5 .6 .7 I.I 1.3 1.5 1.7 1.9 2.1


%Cr2O3 %AI 2 0 3 %CaO
FIG. 16. Major and minor element compositional parameters for cumulus orthopyroxene infieldsection Tl (see
Fig. 1C for location). The upper 37 m is websterite and the lower portion is bronzitite. The resistant and more easily
weathered layers of the medium-scale layering are shown in relation to the sample positions. Although the
boundaries between these layer types are shown as sharp breaks they are really gradational.
638 ALLAN H. WILSON

increases through the lower bronzitites but shows considerable variation and is strongly
antipathetic to mg-number. Calcium in orthopyroxene increases gradually through the
sequence with small variations; Cr follows closely the trend of wig-number but the small-
scale variations tend to peak at positions slightly below those for mg-number; Ni and Al
have trends which closely follow mg-number.
Field section Tl (located on the east side of the longitudinal axis) does not contain the
small-scale and centimetre-scale layering observed on the west side of the axis in the same
area (Fig. 1C). The upper bronzitites and websterites which exhibit this layering were
intersected in borehole BH1 (Fig. 13). The homogeneous lower bronzitites in this section,
which do not contain fine-scale layering, show cyclical variation of pyroxene compositions
for wg-number (Fig. 17) and minor components. The changes are not marked and are

Downloaded from http://petrology.oxfordjournals.org/ at Aston University on January 30, 2014


resolvable only because of the high precision of the analytical data. In general terms, the mg-
number decreases from ~0-865 at 127 m to 0-835 at 75 m. The pyroxenes then become
overall more magnesian upward in the sequence, this change coinciding with the 2 m thick
olivine bronzitite at 61 m. Minor amounts of olivine are observed at 80 and 109 m, which
also correspond to small peaks in wig-number. Variations of Ni and Cr contents in bronzite
(Fig. 17) correlate closely with the small variations in mg-number, and Ti remains essentially

Websterlte Finely layered bronzitite


I Olivine bronzitite Bronzltite f•'-'.'.''I Olivine-bearing bronzliile

30-

40-

5O-

60-

7O-

8O-

9O-

IOO-

110-

120-
I • r \ ~!~" ' I I ' I • I ' I
.83 .84 .85 .86 .06 .08 .10 .5 .6 .08 .10 .12 .14 1.3 1.5 1.6 2.0
Mg/(Mg+Fe2») %NIO %Cr 2 O 3 % TiO 2 %CoO

FIG. 17. Major and minor element compositional parameters for cumulus orthopyroxene in borehole BH 1. Rock-
types are shown together with stratigraphic division into Subunits la and lb. Compositions of olivine (Fo) are also
given. Compositions are shown for the finer-grained (open dots) and coarser-grained (solid dots) layers at the top of
the bronzitite succession. The dashed lines highlight positions of small compositional reversals for Mg/(Mg
+ Fe 2+ ). The positions of the resistant (open bar) and more easily weathered layers (black bar), as extrapolated from
field observations, correspond generally to zones of lower-magnesian and more-magnesian pyroxenes respectively.
THE GEOLOGY OF THE GREAT DYKE, ZIMBABWE 639

constant up to the olivine bronzite layer and then decreases. This component also shows
small-scale antipathetic variations with changes in mg-number. Aluminium and Ca both
increase upward with detailed variation that can be correlated with mg-number.
The part of the bronzitite sequence from 25 to 36 m is marked by alternation on a small
scale of coarser- and finer-grained layers, which continue upward into the websterite. In the
upper 36 m of the PI succession, 19 pairs of the two-layer types are present, and lithological
and mineral compositional data are shown in detail in Fig. 18. The small-scale cyclic
variations in mg-number in both the alternating grain-size layers are superimposed on the
lithological and modal changes which occur over much smaller vertical intervals than the
cryptic variations. The alternating, poorly orientated coarser-grained, and well-oriented
finer-grained layers thus represent a fine structure within the broader-scale compositional

Downloaded from http://petrology.oxfordjournals.org/ at Aston University on January 30, 2014


variation. The pyroxene compositions of the finer-grained layers are significantly more
magnesium rich than those of the coarser-grained layers. The small-scale layers within the
broader compositional variations are therefore identified by compositional differences as
well as by significant petrological parameters such as grain shape, size, and mineral fabric. A
very significant observation is that the distribution of small-scale layers in the websterite
unit is closely related to the broader compositional cryptic variations (Fig. 18) in that the
finer-grained layers tend to be more frequent where the pyroxenes in both layer types are
more magnesian (for example, pyroxenes in the sparsely layered intervals 9-14 m and
18-21 m have relatively low mg-numbers). Such a complex interplay of mineral composi-
tions, textural types, and lithologies has not previously been described in layered intrusions.
The minor elements display variations clearly identified with the two layer types (Fig. 18).
Chromium shows variations compatible with mg-number whereas Ti is antipathetic.
Surprisingly, Ni is significantly lower in the finer-grained pyroxenes even though these are
more magnesian. In both layer types this element tends to show an antipathetic variation in
relation to mg-number. Aluminium and Ca in bronzite show positive correlation with the
cyclic variation in mg-number. Overall, Ni, Cr, and Al in orthopyroxene decrease upwards in
the websterite unit and Ti increases.
Very coarse-grained (up to 20 mm long) bronzite megacrysts (Fig. 10C) associated with
the finer-grained layers are slightly more magnesian than the pyroxenes with which they are
associated, and in some cases considerably richer in Ni. The megacrysts are richer in Cr
relative to the associated pyroxenes, but have similar contents of Ti and Al. A striking
feature of the bronzite megacrysts is their exceptionally high Ca contents, which increase
from 1-8% CaO at 35 m to 2-4% at 14 m (Fig. 18).
Compositions of cumulus augite (not shown) from the websterite unit show cyclicity
similar to that of bronzite for mg-number and minor elements, and also the same
compositional contrasts between the coarser- and finer-grained layers.
Other borehole sections do not exhibit similar small-scale layering but complex composi-
tional variations are observed. Compositional variations of cumulus bronzite are shown for
borehole section BH4 (Fig. 19), located near the west margin (Fig. 1C). Relative variations
of mg-number and minor elements are similar to that described for BH1, and several
compositional reversals are present.

Lateral variations of pyroxene compositions


Significant lateral variations of the major and minor element compositions for cumulus
pyroxenes are observed in the Darwendale Subchamber. The general tendency from axis to
margin is for pyroxenes to have lower mg-number, Cr, and Ni, and higher Al, Ti, and Ca.
ed from http://petrology.oxfordjournals.org/ at Aston University on January 30, 2014
o Coarser-grained Finer-grained x - Megocryst
LC >r
0- .-O
2-
C,_ 1
J
o 6-
8
C2-
c3- _J i - ->$ .-* }-$-- -V ^
§ " ct- T- 3r
o la- • t
a

's l214-
01 ~ c5-
<h-• :
c,- F5
r
7
o 16- c«- • -Fe t-
<D •F9
a 18- c9 • O r
,*r
z
|22- u Ik P 3
c24- -

g26- c,r
|28- c,5- 4-^ C/3
O
230- ~ z
Q c
32- J
34-
c- r^«
F
m

r »
36- Si" D

Mg/<Mg+ Fe2+)

FIG. 18. Major and minor element compositional parameters for orthopyroxene through the upper bronzitite and websterite of Subunit la in
borehole section BH 1. This section consists of alternating small-scale finer-grained (F; solid dots) and coarser-grained (C; open dots) layers shown
in Fig. 13. Individual small-scale layers are numbered downwards. Horizontal dashed lines highlight positions of peaks in mg-number of the
pyroxenes. Small crosses ( x ) are compositions of bronzite megacrysts. The hatched zone on the column represents missing core.
THE GEOLOGY OF THE GREAT DYKE, ZIMBABWE 641

Iwebsterite I | Bronzitite
o
s
10
— 15
u
° 20
1 2»
2 301

3 40
O
.n 45"
"5
«, 50-

Downloaded from http://petrology.oxfordjournals.org/ at Aston University on January 30, 2014


v
i 55-
S 60
65
70
J
75
.80 Bl .82 83 84 .3 4 .5 .07 09 II .12 .14 16 18 201.2 14 1.6 1.6 1.8 2.0 2.2
Mg/(Mg»Fe2*) % Cr2O, % NiO % TiO2 % AI 2 0 3 % CaO

FIG. 19. Major and minor element compositional parameters for orthopyroxene through the upper bronzitites and
webstentes of Subunit la and lb of borehole BH4 on the west margin. Horizontal dashed lines highlight positions
of peak values for mg-number.

Smoothed profiles of bronzite compositions (Fig. 20) for the series of borehole sections at
different distances from the axis of the subchamber (Fig. 1C), constructed on a relative basis
for each section to account for the progressive thinning of the PI unit from axis to margin,
show consistency in the position of the major compositional reversal marking the base of
subunit la. Some of the minor fluctuations in composition may also be correlated between
sections but the overall trends are different for each section.
Mg-number and Cr content decrease upwards through a series of oscillations to the 035
position, above which most sections show a reversal to more primitive compositions. This
reversal is gradational and after attaining a maximum starts decreasing again. Titanium
remains effectively constant over this section with large fluctuations. There is a reversal to
more magnesian compositions and lower Ti contents just above or below the
websterite-bronzitite contact, followed by a change to less magnesian pyroxene composi-
tions together with an increase in Ti content. The variation of Ti is the inverse of that of mg-
number. Chromium does not display the same variation but instead decreases (by a factor of
three) from the base of the websterite unit and shows a small inflection near the centre of the
unit.
The overall decrease in orthopyroxene mg-number both upward in the succession and
laterally is shown in the three-dimensional drawing (Fig. 21) of the Darwendale Sub-
chamber. Contours were determined for each of the boreholes and field sections by
averaging pyroxene compositional data in groups as representing seven compositional
intervals through the layer to emphasize the broad compositional trends. Discordance
between the modal layering (as represented by the PI package) and cryptic variation is
apparent. Lateral asymmetry is also observed, with the pyroxenes in the west marginal
facies being considerably less magnesian than those of the east facies at the equivalent
stratigraphic position.
ded from http://petrology.oxfordjournals.org/ at Aston University on January 30, 2014
Borehole sections (BH)
Sal.

t-
r"
Z
X

r
O
Z

7 8 .79 80 81 82 83 84 85 86 .2 .3 4 .5 .06 10 .12 14 16 .18 .20


Mg /(Mg->Fe**) Cr2O3 % TiOj,
FlG. 20. Summary of pyroxene compositional parameters (mg-number, %Cr2O3, %TiO2) from borehole sections (BH number given and
positions shown in Fig. 1C; Sel represents sample from Selous Mine). All stratigraphic sections have been normalized to the (hickness of the
PI layer in the axis, and are therefore represented as the relative proportional distance below the mafic-ultramafic contact. Small-scale
fluctuations and analytical error were reduced by generating the curves using three-point moving averages. The vertical bar represents the
more easily eroded layers (solid) and the more resistant layers (open) in the axial environment of the PI bronzitite and has been extrapolated
fromfielddata (see Figs. 5 and 16). Shaded bands indicate minor compositional reversals and suggest weak correlation between sections. The
solid dots in the gabbro zone indicate the compositions of orthopyroxene in the mafic rocks immediately overlying the websterite layer (these
points are shown above the contact for clarity). The expanded upper scale of Cr2O3 contents applies only to pyroxenes from the gabbroic
rocks.
ed from http://petrology.oxfordjournals.org/ at Aston University on January 30, 2014
H
X
m
0 84, O
0 88 0.86 0 84 m
O
088 r
O
O
•<

o
-n
Selous Mine H
X
.© m
- 0 79 O
0 82 0.83

>
084
H
0 86 D
•<
0 88
m
FIG. 21. Three-dimensional drawing of the PI bronzitite layer in the Darwendale Subchamber, showing orthopyroxene compositional contours N
for mg-number. The contours were established by taking averages for groups of samples to make up the compositional intervals as shown. The
heavy vertical lines mark the positions of the boreholes as shown in plan view in Fig. 1C and the heavy dashed lines mark the field traverses (T).
2
03

f
644 ALLAN H. WILSON

Discontinuities in compositional trends for cumulus orthopyroxene


Bronzite compositional parameters, represented in terms of stratigraphic variation, do
not indicate obvious decoupling of major and minor elements (see Figs. 16-20). However,
when plotted against mg-number, the minor components (Ti, Ca, Al, Cr, and Ni) form a
series of subparallel linear arrays (Fig. 22). Each of the linear trends incorporates all samples
taken within specific intervals in the borehole sections. The Ti variation diagrams show
arrays indicating antipathetic variation (increasing Ti with decreasing mg-number). In each
borehole section the arrays show progressively decreasing mg-number but the ranges in Ti
values are similar. As a group, bronzites from the finer-grained layers in the websterite and
bronzitite (section BH1) are more depleted in Ti than those from the coarser-grained
varieties. Each of the borehole sections shows similar array patterns, and stratigraphic

Downloaded from http://petrology.oxfordjournals.org/ at Aston University on January 30, 2014


intervals represented vary from 10 to 30 m for the lower bronzitites and from 3 to 10 m in
the upper bronzitites and websterites.
Aluminium in bronzite (Fig. 22) shows similar offset arrays but there is greater scatter
than that observed for Ti. The slope of the individual lines within the array is opposite to
that for Ti, with A12O3 varying sympathetically with mg-number. Calcium in bronzite
exhibits well-defined linear arrays, with individual trends showing an increase with mg-
number. Chromium shows a high degree of correlation with mg-number for individual
arrays. The consistent and well-constrained trends reflect the high analytical precision for
this element. Bronzite compositions from the websterite layer are compositionally divided
into two groups, both of which are characteristically depleted in Cr relative to orthopyrox-
ene from the bronzitites. In BH1 the finer- and coarser-grained websterites each comprise
two separate linear arrays.
The variation diagrams for Ni in bronzite also indicate a series of parallel arrays but these
are generally less well constrained than for the other minor elements. Orthopyroxene from
the lower bronzitites shows decreasing Ni content with mg-number but the upper bronziti-
tes in some sections, and all samples from the websterite layer, show increasing Ni with
decreasing mg-number. Stratigraphic intervals defined by the Ni arrays are also commonly
different from those of the other minor elements, particularly towards the top of the PI
bronzitite, indicating other controls (most likely the presence of sulphide) on the behaviour
of this component.
Superimposition of the linear arrays for each of the elements (Fig. 23) exhibits broad
trends. Titanium shows subparallel and closely overlapping trends, with significant varia-
tion in Ti at any given mg-number. The arrays for the bronzitites are generally slightly
steeper than those for the websterites, but the overall trend is one of increasing TiO 2 with
decreasing mg-number. Aluminium (Fig. 23B) shows an overall increase with decreasing
mg-number but the antipathetic variation for each array is highly consistent. Calcium
(Fig. 23C) shows similar trends to Al, except for the most magnesian pyroxenes which have
a very small range in Ca contents. The general trend of sympathetically decreasing Cr with
mg-number (Fig. 23D) is apparent. Combining the arrays for Ni (Fig. 23E), the overall trend
is one of sympathetically decreasing Ni with mg-number of bronzite; however, for any given
mg-number the Ni content varies widely. Compositions of orthopyroxene in the gabbro
overlying the websterite layer are also shown in Fig. 23 for comparative purposes.
The pyroxene composition variation diagrams for minor element components with mg-
number identify discrete trends in all the sections studied. The variations are not of a simple
cyclic nature and there is no systematic progression of compositions with stratigraphic
height within each of the units defined in the arrays, nor is there obvious correlation
between borehole sections. In some sections, arrays are displaced from the normal
ded from http://petrology.oxfordjournals.org/ at Aston University on January 30, 2014
0.20-
0.18 -
28 4-37.5-0 V '4»*~>
J-0.16 -
• - 0.14 - m»el>
eg
<^&% . 0
e o.,2 - 37.5-31.2

5? 0.10 -
BH1 BH5I BH31
0.08
ei .82 83 84 85 86 .87.80 .81 .82 .83 84 80 .82 .83 84 X
m
O
0.6- 35J-7OJ broni 70J-I03.0 ^ ° r
m
O
c
M O
o O
•<
c 0.4 -
0-171 f i n , w «b
o
19 2-26.4 web
O 0.3 -
H
CM
<-> 0.2 -
-
0-19.2 ««b
a
O.I BHil BlT3| rn
.81 Sh .85 .86 .87 80 81 82 83 .80 .81 84 a
TO
46.0-62.S m
16 fins bronz 0 09. >
*/ . .381-76,2 2 3-
236-MJ^ 6.0-a i -i
1.5 . 1* ^ 6.0-9.0 |2 0-38.0 X \
i Opx

\ \K 76 2-J03 0
D
1.4 .
1.3 . o 5^o O"0 08- m
o ^ r£tt$f"x ' 0 3 °- 1 2 6 8 = 1 9

1.2 - o 90-iao VM N
I.I -
0-23.6
cri. web.
C
T ^
0-236
(kie wab.
BHli
%_ 0.07-
0-6.0 web ~~
s
03
1.0 BH11 1.5 >
i • i 0 06-
.81 .82 .83 .84 .85 .86 87 .79 80 .81 82 83 84 .81 82 83 84 03
79 .80
m
Mg /(Mg + Fe2-) in Orthopyroxene

FIG. 22. Variation of mg-number with minor components (TiO 2 , A12O3, CaO, Cr 2 O 3 , NiO) in cumulus orthopyroxene in several borehole
sections (BH). For each section, common symbols and regression lines [using the method of reduced major axes after Miller & Kahn (1962)]
define compositional arrays, with numbers referring to the stratigraphic intervals (metres) represented by the samples. (Note that the order of
individual samples within each of the arrays is generally not the order in which they occur in the stratigraphic section because of compositional
variations within the intervals.) These arrays represent a type of cryptic layering not previously recognized in layered intrusions, which is here
called "cryptorhythmic layering". Abbreviations: Bron—bronzitite; Web—websterite; fine—finer-grained layer in BH1; crs—coarser-grained
layer in BH1. Error bars showing precision of analysis are circled.
646 ALLAN H. WILSON

.22 1.8
.20 1.7
© ©
- .16
2 .14 H

-12
.10
1.2-
.08
I.I
1.0
2.3-,

Downloaded from http://petrology.oxfordjournals.org/ at Aston University on January 30, 2014


.10-,

1.5 .09-

7- O .08-

.6 -
O
7 07 H

O .3 - .06-

<-> . 2 - © © ©
I .05 I I I I I I
.7 7 .79 .81 .83 .85 .87 .77 .79 .81 .83 .85 .87
Mg/(Mg*Fe*) in orthopyroxene

FIG. 23. Compositional arrays for cryptorhythmic layering (expressed as percentage of minor element components
against mg-number for orthopyroxene) for all borehole sections combined. (A) TiO 2 ; (B) AI 2 O 3 ; (C) CaO; (D)
Cr 2 O 3 ; (E) NiO. Solid lines represent bronzitites; dashed lines represent websterites. The overall trends are those
expected during fractionation but individual array lines in the CaO, A12O3 plots, and for some in the NiO plot,
show converse relationships (see text for discussion). Circled numbers are for orthopyroxene from the gabbroic
rocks immediately overlying the websterite layer for the various BH boreholes (numbers identify BH boreholes in
Fig. 1C). S—Selous Mine gabbro; *—field sample ofgabbro near axis.

progression and some have slightly different slopes. In a given section, the breaks in the
arrays for each element occur at similar stratigraphic levels although the exact position may
be displaced by one or two samples. This type of cryptic layering based on the covariant
behaviour of major and minor compositional components in orthopyroxene has not
previously been recognized in layered intrusions and is here informally termed 'crypto-
rhythmic layering'. Some of the data suggest that these discontinuities also occur as subsets
within the larger-scale units but cannot be resolved on the present sampling scale.
Lithological changes involving grain-size and modal variations which define the medium-
scale layering may be correlated with discontinuities in the cryptorhythmic layering (Figs. 12
and 13). These, in turn, may be related to the field expression of the layering pattern observed
in the axis of the Darwendale Subchamber.
The most significant observations regarding the cryptorhythmic layering are as follows:
(1) In the present data set there is no discernible correlation of cryptorhythmic layers
between different sections.
THE GEOLOGY OF THE GREAT DYKE, ZIMBABWE 647

(2) Each section studied as a whole is chemically distinctive.


(3) For any given major element composition (in terms of mg-number) for orthopyroxene
there is a wide range in the contents of minor elements.
(4) Significant compositional reversals in the vertical section are also identified by
discontinuities in the cryptorhythmic layering, which in many cases also coincides with the
boundaries of the medium-scale layering.

Modification of cumulus pyroxene compositions


A plot of cumulus orthopyroxene compositions against whole-rock P contents (Fig. 24)
reveals strong linearity for groups of samples identified as belonging to individual
cryptorhythmic units in the different borehole sections. Regression lines through the data

Downloaded from http://petrology.oxfordjournals.org/ at Aston University on January 30, 2014


points in the axial borehole (Fig. 24A) are essentially parallel and show different intercepts
on the compositional axis (mg-number) for each cryptorhythmic array. The cryptorhythmic
arrays in the more marginal section (BH4 in Fig. 24B) have steeper slopes and a very narrow
range of intercepts compared with those from the axial zone. A synthesis for all sections
(Fig. 24C) shows a pattern of progressively decreasing m^-number for orthopyroxene with
increasing P content in the whole rock, but with a general increase in slope for borehole
sections nearer the margins.
The explanation of the trends would be that the variation of mg-number for orthopyrox-
ene in each of the cryptorhythmic units is related to the amount of trapped intercumulus
liquid which has reacted with the pyroxene, and this amount of liquid is reflected in the
concentration of P measured in this study. Each of the cryptorhythmic units represents

Crs websterile • 103-127m o 58 - 74m


i Finer-grained ° 70-103m * 42-58m
webst. and bronz A
27 - 70m • l2-42m
D
25-27m 6-l2m

0.81 0.82 0.83 0.84 0 85 0.86 0 87 0.79 0 80 0.81 0.82 0.83 0.84 0.85
Tl field section
BHI

0.79 0.80 0.81 0.82 0.83 0.84 0.85 0.86 0.87 0.88
Mg/(Mg*Fe2*) In orthopyroxene

FIG. 24. Plots of P vs. ma-number for orthopyroxene, showing offset arrays. Regression lines for data sets are
shown. (A) Relationship for borehole BHI. Finer-grained bronzitite and websterite have very low P contents or are
at the limit of detection. (Note the wide range of intercepts for the arrays on the m^-number axis.) (B) Relationship
for borehole BH4. Compared with BHI the intercepts on the mg-number axis are very restricted. In this plot, some
points are displaced from the regression lines (symbols shown with ticks). (C) Summary of data for borehole sections
and field section T l . Ranges of P content and slopes of lines increase progressively from the axial to marginal facies.
648 ALLAN H. WILSON

crystallization from a discrete magma layer and the difference in slope would be related to
the relative amount of P in the magma. The liquids which contained more P would have
steeper slopes. The intercept on the compositional axis (at zero P content) would define the
liquidus composition of pyroxene which would have crystallized as a perfect adcumulate, i.e.,
one in which there had been no trapped liquid and therefore the orthocumulus component
would have been zero. These data provide a unique opportunity to estimate the composi-
tional changes that took place by interaction with the trapped liquid.

Modelling of cumulus processes and determination of residual porosities


As the final compositions of cumulus pyroxene and the modal percentage of cumulus

Downloaded from http://petrology.oxfordjournals.org/ at Aston University on January 30, 2014


pyroxene are known, and the initial compositions can be determined from the mg-number vs.
P plot, it is possible, on the basis of mass balance calculations, to estimate the residual
porosity of the rock and the amount by which the liquidus pyroxenes must have been
enlarged by growth from the trapped liquid. It is not necessary to know the P content of the
liquid but an estimate is required of the MgO content of the parent magma. In the case of the
PI layer, the parent liquid composition (in particular, the MgO content and Fe/Mg) is taken
from Wilson (1982) as that being appropriate to the base of PI, and then adjusted to be in
equilibrium with the determined liquidus composition (given by the intercept value) by
fractional crystallization of orthopyroxene. Mass balance is achieved on the basis that all the
Mg is used up from the trapped liquid to form zoned orthopyroxene at the same instant the
final porosity is reduced to zero. Reaction between the growing solid and the evolving
trapped liquid would also have taken place, but the end-result of the processes of
solid-liquid reaction and overgrowth would have been the same. The method of calculation
(for details see Appendix) is similar to that used by Barnes (1986), except that in the present
case the original pyroxene composition is known and the residual porosity is calculated,
whereas the porosity has to be estimated from the amount of postcumulus material in the
approach used by Barnes. It has already been shown that in the Great Dyke the amount of
postcumulus phases is not a reliable estimate of the trapped liquid. In addition, the model
evaluates the mineralogical budget of postcumulus overgrowth on the cumulus orthopyrox-
ene, the calculated porosity, and the observed amount of discrete postcumulus material, and
determines the heteradcumulus component of the oikocrysts. The result is relatively
insensitive to fairly large errors in the composition of the parent liquid but is strongly
dependent on the liquidus and final compositions of the cumulus orthopyroxene. Both these
parameters are accurately determined.
Data from five cryptorhythmic layers have been taken as examples of the modelling; two
are from the axial section of borehole BH1, and three from the marginal environment (BH4,
BH5, and BH7). Plots of mg-number vs. whole-rock P content (Fig. 25) permit the liquidus
pyroxene compositions for each group of samples to be determined. The criterion used for
establishing a collinear array of data points is that > 75% of all samples must lie within three
times the analytical error of the line representing the linear correlation of the data points.
Samples for which data points lie outside this limit are regarded as having undergone
secondary processes of addition or removal of trapped liquid. In the Great Dyke the degree
of scatter becomes progressively greater for the marginal facies, and four points in the array
for borehole BH7 (Fig. 25) are significantly displaced from the array.
The residua] porosity as determined in the model is shown plotted against observed mg-
number (Fig. 26A) for each of the pyroxenes and against P content of the whole rock
(Fig. 26B). The residual porosities for each of the sections studied are similar irrespective of
THE GEOLOGY OF THE GREAT DYKE, ZIMBABWE 649

200-,
180 -
160 - BH7( 8-l4m)

t 140 -
E
o 120 H
o
"I 100 -

~ 80-
Q. 0.80 0.81 0 82 0.83
3- 60-

Downloaded from http://petrology.oxfordjournals.org/ at Aston University on January 30, 2014


40-
20-
0
0.82 0.83 0 84 0.85 0.86
z
Mg / (Mg+Fe *) in orthopyroxene

FIG. 25. Whole-rock phosphorus vs. mg-number of orthopyroxene for several sections studied in detail and used in
modelling (see Appendix). BH numbers refer to borehole and numbers in parentheses give the stratigraphic
intervals. Error bars are circled. Intercepts on the mg-number axis for each array are considered to give the liquidus
compositions before modification of the pyroxene by postcumulus processes. Data points for BH7 which are ticked
represent samples which are significantly displaced from the regression line.

the positions of the boreholes relative to the margin or axis of the subchamber. This indicates
that a high proportion of the discrete postcumulus phases (interstitial clinopyroxene and
plagioclase) did not form from the trapped liquid and therefore grew as heteradcumulus
oikocryst phases. This conclusion is consistent with petrographic observations (particularly
regarding the nodular pyroxenite) in which a proportion of the discrete postcumulus phases
is considered to have formed at a relatively early stage. The results of the modelling are
shown in Table 2. The change of slopes in the plot of porosity vs. whole-rock P content
(Fig. 26B) also reflects the different slopes in the mg-number vs. P plots observed in Fig. 25.
This means that primary magma in the marginal sections had significantly more P than the
magma in the axial environment by up to a factor of three. It is also clear that the more iron-
rich pyroxene composition of the marginal facies is a. primary feature related to the
composition of the magma and cannot be explained entirely by the role of the trapped liquid.

Origin of the cryptorhythmic units


Data points defining the cryptorhythmic units are the same as those in the individual
offset arrays of cumulus orthopyroxene composition plotted against whole-rock P content.
The displaced trends also indicate that there is not a single evolutionary path of the magma,
and this points to primary compositional heterogeneity within the magma body. It must
therefore be concluded that the variation of minor components in pyroxene within each
cryptorhythmic unit reflects the interaction of the pyroxene with the trapped liquid during
the solidification process.
ded from http://petrology.oxfordjournals.org/ at Aston University on January 30, 2014
16 i
14 •

12 •
10 BHI U03-l27m)
8
6.
g 4 •

2-

Q- 0.80 0.81 0.82 0.83 0.84 0.85 0 86 0.87


"5
| 16-, Mg /(Mg«Fe 2 *) for orthopyroxene
en
£ 14-1 z
1 l2- BH7 X
1 10-
r
u 8- to
6- O
4-
z
2-

20 40 60 80 100 120 140 160 ISO 200 220 240 260 280 300
P (ppm) in whole rock

FIG. 26. Results of modelling of residual porosities which would cause the observed displacement from the original liquidus
compositions. (A) Residual porosity vs. mg-number for the sections shown in Fig. 25. (Note that in all sections some samples
are indicated to have had very low porosities.) (B) The same data sets shown plotted against P content in the whole rock. The
marked change in slope indicates that the magma in the marginal facies had higher P content than that in the axis of the
subchamber.
THE GEOLOGY OF THE GREAT DYKE, ZIMBABWE 651

PETROGENESIS AND MAGMATIC EVOLUTION OF THE PI


LAYER OF THE GREAT DYKE
Origin of the cumulates
Liquidus compositions of pyroxenes in the PI layer of the Great Dyke have undergone
extensive modification as a result of interaction with trapped liquid. There is also a
progressive change in liquidus pyroxene compositions from axis to margin, indicating a
primary lateral control in the magma chamber. This is supported by evidence that the P
content of the magma was significantly greater in sections closer to the margins. The
marginal facies is characterized by greater amounts of discrete postcumulus phases relative
to the axis, but numerical modelling of the crystallization and re-equilibration process shows
that overall the residual porosity increases only slightly from axis to margin (Table 2), but

Downloaded from http://petrology.oxfordjournals.org/ at Aston University on January 30, 2014


with some cumulates from the marginal facies showing very low residual porosity. There is
evidence that migration of trapped liquid took place in some cumulates from the marginal
zone, resulting in depletion or addition of incompatible trace elements. Migration of trapped
liquid was less in the axis, resulting in very coherent data sets. Cumulates characterized by a
relatively open network of crystals (i.e., with higher porosity) have a greater chance of
undergoing migration of the trapped liquid. In the case of the PI layer, cumulates having a
porosity < 10% have not experienced liquid migration whereas at higher porosities, and
particularly for rocks in the marginal facies, migration of liquid is indicated.
A high proportion of the discrete postcumulus phases grew in equilibrium with the parent
magma or in a steady-state situation with magma that was only slightly depleted. This
steady-state situation may be envisaged in which evolved magma derived from the
environment of the growing crystals was partly removed in the open system and replaced or
mixed with undepleted parent magma by a convective process similar to that suggested by
Tait et al. (1984). These phases (plagioclase and clinopyroxene in the bronzitites and
plagioclase in the websterites) formed oikocrysts and are regarded as having a high
heteradcumulus component. The sequence of crystallization and re-equilibration taking
place at the zone of crystallization is summarized in Fig. 27. The relative development of
oikocrysts (and the formation of the nodular pyroxenite) in the marginal and axial facies is a
result of the contrasting thermal profiles in the two different environments (Fig. 28).
Campbell (1987) noted that adcumulus and orthocumulus formation may be related to the
thermal gradient in the crystallization zone. Crystallization of any phase occurs when the
crystallization temperature (which is usually below the Hquidus temperature) is less than the
actual temperature of the magma. The two situations shown in Fig. 28 correspond to an
environment of high heat flow through the floor as exists in the margin of the Great Dyke,
and to a condition of low heat flow as would occur in the axis. Where the heat flow is high
there is a likelihood that discrete postcumulus phases (augite and plagioclase) would
nucleate and grow while the crystal-liquid system was still open and able to interact with the
overlying body of magma. In this case, a high proportion of these phases would have been
derived from the parent magma and therefore have a high adcumulus component. Once the
system was closed the trapped liquid solidified relatively quickly because of the high thermal
gradient. In contrast, cumulates in the axial environment would have formed at a slower rate
and therefore the system would have become closed nearer to the interface with the body of
parent magma. Most of the intercumulus liquid would have been excluded by the growing
adcumulus orthopyroxene crystals. Discrete postcumulus phases would have nucleated in
the closed system low down in the crystal pile and therefore would have grown almost
entirely from the trapped liquid. In this case, these phases would represent a relatively high
proportion of orthocumulus component.
ed from http://petrology.oxfordjournals.org/ at Aston University on January 30, 2014
TABLE 2
Modelled cumulus parameters for selected cryptorhythmic units

Borehole section
BH1 BH4 BH5 BH7

Interval in section (m) 103-127 24-42 24-33 8-14 z


% Discrete postcumulus phases 1-6-9-6 (5-2) 13-8-21-5 (18 8) 180-260 (21-8) 26-2-29-7 (27-7) X
Opx liquidus composition (mg-number) 0-864 0-841 0-835 0-824
% Heteradcumulus phases 00-3-9 (21) 12-9-17-2 (17-8) 170-19-5 (18-2) 21-2-26-7 (24-1) r
% Residual porosity 0-3-70 (3-9) 1-2—10-1 (4-9) 1-7-10-2 (5-6) 0-6-131 (6-7)
% Overgrowth on cumulus Opx 0-1-4-4 (20) 0-5^-5 (21) 0-7-2-8 (2-7) 0-3-61 (31) O

The percentage of discrete postcumulus phases is based on modal analyses. Orthopyroxene liquidus compositions are from Fig. 25.
z
Range of values is given for each parameter, with mean in parentheses.
All calculations are on a weight % basis, using the following densities: orthopyroxene 3-2, clinopyroxene 30; plagioclase 2-8; liquid 2-65.
Main body of magma

Nucleation of Opx.

Adcumulus growth of Opx and


nucleation of Cpx.

Liquation of sulphide
Nucleation of plagioclase
Postcumulus overgrowth to
form zoned cumulus Opx
and Cpx oikocrysts. Mutual

Downloaded from http://petrology.oxfordjournals.org/ at Aston University on January 30, 2014


interference of crystals.

Growth of plagioclase oikocrysts.


Redistribution of trapped silicate
liquid and sulphide liquid.

Crystallization of postcumulus
Cpx.

Solidification of sulphide liquid


and crystallization of the late - stage
minerals.

Subsolidus reaction of sulphide


and redistribution by reaction
with trapped fluid phase.

Melt Cpx oikocryst


Unzoned Opx and Cpx crystals Sulphide

Zoned Opx crystal Plagioclase

Unzoned Opx crystal Phlogopite / Postcumulus Cpx

ORDER OF CRYSTALLIZATION
Opx. Cpx » Plag ^ . Phlogopite • Quartz K - feldspar

Sulphide liquation
FIG. 27. Profile through the zone of crystallization and solidification in the PI pyroxenite at the level of sulphide
enrichment in the marginal environment. The profile is divided into four domains. (A) This part of the zone is an
open system and is in perfect equilibrium with the overlying body of magma. Nucleation of orthopyroxene occurs,
followed by that of clinopyroxene. Liquation of sulphide also occurs because this phase had reached saturation in
the silicate magma. (B) This zone is still interactively open with the main body of magma but is not in perfect
equilibrium in that evolved intercumulus liquid from the growing crystals is only partly removed (largely by
convection) and replaced by undepleted magma. Plagioclase develops large postcumulus oikocrysts together with
clinopyroxene, and postcumulus overgrowth of cumulus orthopyroxene produces zoned crystals. Trapped silicate
liquid and sulphide liquid are redistributed about the plagioclase oikocrysts. (C) The system is closed and at the
point of closure the proportion of liquid to solid is referred to as the residual porosity. Separate crystals of
postcumulus clinopyroxene form, and all cumulus and oikocryst phases have strongly zoned overgrowths as the
trapped liquid becomes highly evolved. The assemblage of late-stage minerals (phlogopite, K-feldspar, and quartz)
forms at this stage. Sulphide liquid is redistributed further around plagioclase and clinopyroxene oikocysts. (D)
Crystallization is complete. Solid-state re-equilibration begins in ferromagnesium minerals to produce unzoned
crystals as the rock cools. Plagioclase remains zoned because of its low interdifrusion rates. Hydrous pockets remain
and reaction with sulphide takes place, causing redistribution of sulphide and platinum group minerals together
with alteration of pyroxene. The entire sequence of events advances as crystallization takes place.
654 ALLAN H. WILSON

CRYSTALLIZATION TEMPERATURE

WEBSTERITE GABBRO

OPEN SYSTEM

Downloaded from http://petrology.oxfordjournals.org/ at Aston University on January 30, 2014


OPEN SYSTEM

CLOSED SYSTEM
CLOSED SYSTEM

[ A ] HIGH HEAT LOSS (MARGIN) LOW HEAT LOSS (AXIS)

FIG. 28. Thermal conditions during the formation of the cumulus sequence in the PI layer in the marginal and
axial zones. The relative proportions of adcumulus and orthocumulus components at any stage are shown by the
solid blocks. In an open system the adcumulus component of crystal growth decreases downwards as interaction
with the overlying body of magma diminishes. In a closed system crystal growth is from the trapped liquid and
therefore is entirely orthocumulus. (A) The condition of high heat loss through the floor (marginal facies) results in
high degrees of undercooling and high nucleation rates, leading to crystallization of oikocrysts of clinopyroxene
and plagioclase in an open-system environment. A large proportion of these phases would form by adcumulus
growth (heteradcumulus component). (B) Where heat loss is low (axial facies) the open-system zone is narrow, and
discrete postcumulus phases (clinopyroxene and plagioclase) would nucleate in a closed system and therefore would
form as orthocumulus phases rather than as oikocrysts. The crystallization temperatures for orthopyroxene,
clinopyroxene, and plagioclase are shown to be displaced during bronzitite formation whereas they are coincident
for the two pyroxenes in the websterite (inset C). In the gabbros all three minerals are cumulus and therefore
crystallization temperatures are coincident (inset D).

The net result of these two situations is that the proportions of orthocumulus and
adcumulus components are not markedly different, but that in the marginal environment the
discrete postcumulus phases are largely heteradcumulus whereas they are part of the
orthocumulus assemblage in the axial zone. The uppermost pyroxenites and mafic rocks
would have experienced similar controls. In the websterites (Fig. 28, inset C) both ortho- and
clinopyroxene would have the same crystallization temperature, and in the gabbroic rocks
(Fig. 28, inset D) the two pyroxenes and plagioclase would effectively have crystallized at the
same temperature.

Evolution of the PI layer


Double-diffusive convection and compositionally driven convection are important
processes in fluid-dynamic systems (Turner & Gustafson, 1978, 1981; Huppert & Sparks,
1980; McBirney, 1980; Turner, 1980; Sparks & Huppert, 1984; Turner & Campbell, 1986). A
consequence of these processes is that, in magma chambers, crystallization may occur from
compositionally or thermally zoned magmas. Irvine (1981) and Irvine et al. (1983) first
THE GEOLOGY OF THE GREAT DYKE, ZIMBABWE 655

suggested a radical departure from the traditional concept of floor-accumulation of


cumulates in that crystallization may have occurred by sidewall growth to produce a series
of diachronous layers. Wilson & Larsen (1985) and Wilson et al. (1987) demonstrated
discordant relationships between modal and cryptic layering in the Fongen-Hyllingen
layered mafic complex in Norway, and suggested that crystallization took place on a sloping
surface from a compositionally zoned magma body. Similar discordances have also been
reported in a number of layered intrusions (Weibe & Wilde, 1983; Klemm et al, 1985; Robins
et al, 1987), and now in the present work on the Darwendale Subchamber of the Great
Dyke. Prendergast (1989, 1990) also appealed to a modified Wilson and Larsen model to
explain major lateral variations in lithologies in the Wedza Subchamber of the Great Dyke.
The potential for sidewall crystallization to take place becomes increasingly important
where the width:depth aspect ratio of the magma chamber decreases. In cases where the

Downloaded from http://petrology.oxfordjournals.org/ at Aston University on January 30, 2014


magma chamber is narrow and highly elongate, as for the Great Dyke, the sidewall/floor
influence on the cooling history becomes significant. The characteristic flared shaped
transverse section of the Great Dyke (Prendergast & Wilson, 1989; Wilson & Prendergast,
1989; Wilson et al, 1989; Wilson & Tredoux, 1990), will have a major influence on the
thermal properties of the magma chamber, with significant heat loss taking place from the
wall/floor of the chamber, although the dominant heat loss will still be through the roof. The
strong lateral variations observed in the Great Dyke are considered to have resulted from (1)
major changes in the thermal regime from axis to margin, and (2) crystallization from more
evolved upper magma layers to give the marginal facies. The influence of the sidewalls is
most pronounced at the level of Cyclic Unit 1 (including the PI layer), where, because of the
flared cross-sectional shape of the Great Dyke, this tabular body will directly overlie and be
juxtaposed to the floor/wall over wide areas. This marginal environment contrasts with
occurrences closer to the axis, where the PI layer overlies progressively thicker amounts of
hot ultramafic cumulates.
Causes of compositionally stratified magma in magma chambers remain unclear. Huppert
et al (1986, 1987) have suggested that compositional stratification may be generated
experimentally in homogeneous solutions above an inclined floor and that compositional
contours are discordant to the crystallization front. This conclusion was refuted by Martin
& Campbell (1988), who concluded that compositional convection from inclined floors will
not in itself lead to compositional zonation in the magma. They considered, however, that a
process more likely to lead to zonation in magma chambers is that in which replenishment
by dense more primitive magma occurs where crystallization processes have resulted in the
release of buoyant fluid. Emplacement of the dense magma in the form of a turbulent
fountain (Campbell & Turner, 1986) would lead to the development of a zoned hybrid near
the base. This would break up into a number of double-diffusive convection layers which
would retain the compositional stratification if each of the layers was undergoing crystalliza-
tion simultaneously. Prendergast (1991) suggested that a transverse compositional gradient
was established in the Wedza Subchamber, giving rise to cooler and more evolved magma
nearer the margin.
Cyclic Unit 1 of the Great Dyke is considered to have formed by a major influx of
primitive magma into the chamber (Wilson, 1982; Prendergast & Wilson, 1989; Wilson &
Prendergast, 1989) which had undergone considerable compositional modification through
extensive crystallization of cumulus olivine and bronzite of the underlying cyclic units. The
buoyant and evolved magma released as a result of this crystallization would have collected
near the roof of the chamber (Sparks & Huppert, 1984). Emplacement of a hot, dense magma
as a fountain would have caused the chamber to become zoned in a series of hybrid layers
established within the mixing profile. The new pulse of magma would have had a marked
656 ALLAN H. WILSON

effect on the lower liquid layers but progressively less influence on the layers higher up in the
magma column. Martin & Campbell (1988) noted that this type of layering will persist for
long periods of time and that mixing and overturning will occur only when the density
difference between adjacent layers becomes very small. Campbell (pers. comm., 1990)
suggested that crystallization from each of the cryptorhythmic units took place from a
discrete magma layer and successive incorporation into the overlying layer resulted in the
offset mineral trends. A major influx of magma initiated Cyclic Unit 1 of the Great Dyke,
also giving rise to the economically important chromitite layers (Wilson, 1982; Prendergast,
1987; Prendergast & Wilson, 1989). Chromitite layers formed near the base of the cyclic unit
by a combination of controlling factors (Irvine, 1977; Wilson, 1982; Murck & Campbell,
1986). This was followed by extensive crystallization of olivine in the lower part of the unit.
Several subunits were formed by repeated influxes of primitive magma, resulting in a

Downloaded from http://petrology.oxfordjournals.org/ at Aston University on January 30, 2014


repetition of stratigraphy. Periodic emplacement of magma could have sustained the
stratified magma column established at an earlier stage as well as having important controls
on initiation and cessation of sulphide precipitation (Campbell et al, 1983; Naldrett &
Wilson, 1990; Naldrett et al., 1990). As crystallization of the cyclic unit proceeded,
orthopyroxene joined olivine as a liquidus phase and then became dominant to form the PI
bronzitite in Subunit lb. Crystallization of the entire sequence, now represented by Cyclic
Unit 1, would have taken place against a sloping wall/floor.
The complex interplay of chemical and physical processes related to crystallization and
emplacement of magma resulted in gradual but repeated changes in major element
composition, grain-size, crystal shape, crystal fabric, and distribution of postcumulus
phases. Decoupling of major and minor elements and of incompatible trace elements in
whole rocks indicate that the magma body was compositionally zoned. The discordance of
the primary liquidus compositions in relation to their position in the PI layer indicated that
crystallization was taking place simultaneously from different magma layers in the chamber.
Repeated influxes of magma as fountains, probably close to the axis of the subchamber,
because of their momentum broke through the liquid layers to higher levels in the magma
chamber. The volumes of magma emplaced to give the minor reversals were less than those
that initiated the entire cyclic unit. Bronzitites containing olivine resulted from these
influxes, which were only of limited extent and restricted to the axial environment. Reversals
in mineral compositions and textural types do not coincide with the appearance of olivine,
which indicates that the influx of magma took place gradually, causing the system to be
progressively modified before olivine finally appeared and orthopyroxenes achieved their
most magnesian compositions. Given appreciable momentum, the new influxes would have
risen to a considerable height within the chamber before falling back and affecting the
magma column in several ways. Some degree of mixing would have occurred within the
liquid layers, causing them to become more magnesium-rich in composition, but with
progressively decreasing effect higher in the chamber and also laterally within the layers. The
influence of the mixing process will have been transmitted laterally through each of the
liquid layers and caused a relatively slow reversal in pyroxene compositions at the
crystallization surface. The most pronounced of the reversals gives rise to the boundary
between Subunits la and lb. Increased rate of convection within the liquid layers
consequent upon mixing of the hotter magma would account for changes in rate of crystal
nucleation and crystal shape characteristics. In turn, these crystals would adopt a strong
preferred orientation in the moving fluid regime. Chen & Turner (1980) showed experi-
mentally that double-diffusive layers in the liquid influence the growth of crystals. More
rapid dissipation of the liquid boundary layer to the growing crystals may have influenced
their shape (Martin & Campbell, 1988). The magma fountain would also have helped sustain
the stratified magma column.
THE GEOLOGY OF THE GREAT DYKE, ZIMBABWE 657

The gabbroic rocks overlying the PI websterite layer represent the first appearance of
cumulus plagioclase in the Great Dyke and their emplacement is the last of a series of
complex and interrelated events. Field evidence described above, and for the Wedza
Subchamber (Prendergast, 1991), shows the emplacement to be highly erosive of the
underlying cumulates and suggests that crystallization of plagioclase had already started
elsewhere in the magma chamber (Naldrett & Wilson, 1989,1990; Wilson & Naldrett, 1989).
There is evidence that resorption of pyroxenes in the websterite occurred with the
emplacement of the gabbroic rocks. Preferential erosion of the websterites occurred in
topologically high areas.

Downloaded from http://petrology.oxfordjournals.org/ at Aston University on January 30, 2014


Lateral asymmetry in the Darwendale Subchamber
Major asymmetry is observed between the east and west marginal zones for most
petrological features investigated, with the layered succession on the eastern margin having
characteristics more akin to the axial zone. The origin of this asymmetry may be (1) the result
of tilting of the entire chamber to the west, (2) the deepest part of the chamber being located
east of the geometric axis, or (3) the wall/floor of the magma chamber being more steeply
inclined on the east side than on the west side, resulting in the greater proximity of the
crystallization surface to the country-rock on the west side, with subsequent greater heat
loss. There is no unequivocal evidence to support any of these suggestions but the
topological variations observed on the western margin, as well as the location of the grain-
size layering, which occurs only on the west side of the axis, would tend to support the third
hypothesis.

CONCLUDING REMARKS
Complex lateral and vertical variations in the PI pyroxenite layer of Cyclic Unit 1 of the
Darwendale Subchamber indicate that the formation of the layered sequence was controlled
by both internal and external influences on the magma chamber. A simple model for the
accumulation of crystals at the base of the chamber is not envisaged. Instead, there was a
strong interplay between crystallization from the wall/floor of the chamber, new influxes of
magma, and fluid dynamic processes.
An important feature recognized in the Darwendale Subchamber, and not previously
observed in other layered intrusions, is the superimposition of several layer types over
different vertical intervals. The term 'cryptorhythmic layering' is introduced to describe
sections of the stratigraphy which are uniquely characterized by distinct minor element
compositions for the cumulus pyroxenes. The complex cryptic layering is considered to
reflect small compositional differences in the liquid layers in the magma body established by
double-diffusion convection, each of which imparted its signature on the cumulus crystals
forming from the layers.
The broader-scale cryptic variations are related to periodic influx of new magma,
probably as fountains, which affected all the liquid layers to varying degrees. The results of
these dynamic processes are superimposed on the overall evolutionary trend of the magma.
This section of the Great Dyke stratigraphy is important because it represents the critical
region where plagioclase starts crystallizing, and this too is considered to have been caused
by major perturbations within the system. The sulphide and associated platinum group
element mineralization are also related to these processes.
Major lateral variations in thicknesses of units, mineral compositions, cumulus type, and
mineralization in the Great Dyke (Naldrett & Wilson, 1990; Wilson & Tredoux, 1990;
658 ALLAN H. WILSON

Prendergast, 1991) are related to cooling regimes imposed by the flared structure of the
magma chambers and the decreasing thickness of hot insulating cumulates towards the
margin. There is strong evidence to suggest that the magma giving rise to the marginal zone
rocks was more evolved than that at the axis. This may be the result of a stratified magma
chamber or a pronounced lateral compositional gradient.
The recognition that mineral compositions reflect decoupling of major and minor
elements during fractional crystallization in a magma chamber has important implications
for interpreting binary element plots commonly used in evaluating petrogenesis of lavas.
Discontinuities in such geochemical trends may not reflect different source regions but may
be an expression of the fluid dynamic processes and thermal properties of the magma
chamber.

Downloaded from http://petrology.oxfordjournals.org/ at Aston University on January 30, 2014


ACKNOWLEDGEMENTS
Financial support for this project was provided at various times from 1974 by the
universities of Zimbabwe, Oxford, and Natal, as well as by the Foundation for Research
Development (South Africa). The manuscript greatly benefited by critical comments of T. N.
Irvine, I. A. Campbell, and R. G. Cawthorn. The project was initiated by J. F. Wilson
(University of Zimbabwe), who is thanked for guidance and much support. Union Carbide
(Zimbabwe) is thanked for supporting this project over many years and giving permission to
publish information derived from its drilling programme and borehole cores. X-ray
fluorescence facilities were provided by the Goldfields Laboratory, Geology Department,
University of Natal, Pietermaritzburg.

REFERENCES
Barnes, S. J., 1986. The effect of trapped liquid crystallization on cumulus mineral compositions in layered
intrusions. Contr. Miner. Petrol. 93, 524-31.
Bottinga, Y., & Weill, D. F., 1970. Densities of liquid silicate systems calculated from partial molar volumes of oxide
components. Am. J. Sci. 269, 169-82.
Brothers, R. N., 1964. Petrofabric analyses of Rhum and Skaergaard layered rocks. J. Petrology 5, 255-74.
Buchanan, D. L., 1977. Cryptic variation in minerals from the Bushveld Complex of the Bethal area. Geol. Soc. S.
Afr. Trans. 80, 49-52.
1979. A combined electron microscope and electron microprobe study of Bushveld pyroxenes from the Bethal
area. J. Petrology 20, 327-54.
Campbell, I. H., 1978. Some problems with the cumulus theory. Lithos 11, 311-23.
1987. Distribution of orthocumulate textures in the Jimberlana intrusion. J. Geol. 95, 35-54.
McCall, G. J. H., & Tyrwhitt, D. S., 1970. The Jimberlana Norite, Western Australia—a smaller analogue of
the Great Dyke of Rhodesia. Geol. Mag. 107, 1-12.
Naldrett, A. J., & Barnes, S. J., 1983. A model for the origin of the platinum-rich sulphide horizons in the
Bushveld and Stillwater Complexes. J. Petrology 24, 133-85.
Turner, J. S., 1986. The influence of viscosity on fountains in magma chambers. Ibid. 27, 1-30.
Cawthorn, R. G., McCarthy, T. S., & Da vies, G., 1983. Vertical chemical gradients in a single grain of magnetite
from the Bushveld Complex, South Africa. Miner. Mag. 47, 27-34.
Chen, C. F., & Turner, J. S., 1980. Crystallization in a double-diffusive system. J. Geophys. Res. 85, 2573-93.
Donaldson, C. H., 1974. Olivine crystal types in harrisitic rocks of the Rhum pluton and Archean spinifex. Geol. Soc.
Am. Bull. 85, 1721-6.
Gough, D. I., & van Niekerk, C. B., 1959. A study of the palaeomagnetism of the Bushveld gabbro. Philos. Mag. 4,
126-36.
Hamilton, J., 1977. Sr isotope and trace element studies of the Great Dyke and Bushveld mafic phase and their
relation to early Proterozoic magma genesis in southern Africa. J. Petrology 18, 24-52.
Henderson, P., 1975. Geochemical indicator of the efficiency of fractionation of the Skaergaard intrusion, east
Greenland. Miner. Mag. 40, 285-91.
Hess, H. H., 1939. Extreme fractional crystallization of basaltic magma: the Stillwater igneous complex. Trans. Am.
Geophys. Union 20, 430-2.
Huppert, H. E , & Sparks, R. S., 1980. The fluid dynamics of a basaltic magma chamber replenished by influx of hot,
dense, ultrabasic magma. Contr. Miner. Petrol. 75, 279-89.
THE GEOLOGY OF THE GREAT DYKE, ZIMBABWE 659

Whitehead, J. A., & Hallworth, M. A., 1986. Replenishment of magma chambers by light inputs. J.
Geophys. Res. 91, 6113-22.
Wilson, J. R., Hallworth, M. A., & Leitch, A. M., 1987. Laboratory experiments with aqueous solutions
modelling magma chamber processes II. Cooling and crystallization along inclined planes. In: Parsons, I. (ed.)
Origin of Igneous Layering. NATO AS1 Series. Dordrecht: Reidel, 539-68.
Irvine, T. N., 1977. Origin of chromitite layers in the Muskox intrusion and other layered intrusions: a new
interpretation. Geology 5, 273-7.
1980. Magmatic infiltration metasomatism, double-diffusive fractional crystallization, and adcumulus growth
in the Muskox Intrusion and other layered intrusions. In: Hargraves, R. B. (ed.) Physics of Magmatic Processes.
Princeton, NJ: Princeton University Press, 325-84.
1981. A liquid-density controlled model for chromitite formation in the Muskox Intrusion. Carnegie Inst.
Wash. Yearb. 80, 317-24.
1982. Terminology for layered intrusions. J. Petrology 23, 127-62.
Keith, D. W., & Todd, S. G., 1983. The J-M platinum-palladium reef of the Stillwater Complex, Montana: I.
Origin by double-diffusive convective magma mixing and implications for the Bushveld Complex. Econ. Geol.

Downloaded from http://petrology.oxfordjournals.org/ at Aston University on January 30, 2014


78, 1287-334.
Jackson, E. D., 1961. Primary textures and mineral associations in the Ultramafic Zone of the Stillwater Complex,
Montana. U.S. Geol. Surv. Prof. Paper 358, 106 pp.
1967. Ultramafic cumulates in the Stillwater, Great Dyke and Bushveld intrusions. In: Wyllie, P. J. (ed.)
Ultramafic and Related Rocks. New York: John Wiley, 20-38.
1970. The cyclic unit in layered intrusions—a comparison of repetitive stratigraphy in the ultramafic parts of
the Stillwater, Muskox, Great Dyke and Bushveld Complexes. Spec. Publ. Geol. Soc. S. Afr. 1, 391-424.
1971. The origin of ultramafic rocks by cumulus processes. Fortschr. Miner. 48, 128—74.
Klemm, D. D., Ketterer, S., Riechhardt, F., Steindl, J., & Weber-Diefenbach, K., 1985. Implications of vertical and
lateral compositional variations across the Pyroxene Marker and its associated rocks in the upper part of the
Main Zone of the Eastern Bushveld Complex. Econ. Geol. 80, 1007-15.
Lofgren, G. E., & Donaldson, C. H., 1975. Curved, branching crystals and differentiation in comb-layered rocks.
Contr. Miner. Petrol. 49, 309-19.
Martin, D., & Campbell, I. H., 1988. Laboratory modelling of convection in magma chambers: crystallization
against sloping walls. J. Geophys. Res. 93(B7), 7974-88.
McBirney, A. R., 1980. Mixing and unmixing of magmas. J. Volcanol. Geothermal Res. 7, 357-71.
Noyes, R. M., 1979. Crystallization and layering of the Skaergaard intrusion. J. Petrology 20, 487-554.
McElhinny, M. W., & Gough, D. I., 1963. The palaeomagnetism of the Great Dyke of Southern Rhodesia. Geophys.
J. 7, 287-303.
McKenzie, D., 1984. The generation and compaction of partially molten rock. J. Petrology 25, 713-65.
Miller, R. L., & Kahn, J. S., 1962. Statistical Analysis in the Geological Sciences. New York: John Wiley, 483 pp.
Morse, S. A., 1979. Kiglapait geochemistry II: petrography. J. Petrology 20, 591-624.
1986. Convection in aid of adcumulus growth. J. Petrology 27, 1183-214.
Murck, B. W., & Campbell, I. H., 1986. The effects of temperature, oxygen fugacity and melt composition on the
behaviour of chromium in basic and ultrabasic melts. Geochim. Cosmochim. Ada 50, 1871-87.
Naldrett, A. J., Brugmann, G. E., & Wilson, A. H., 1990. Models for the concentration of PGE in layered intrusions.
Can. Miner. 28, 389-408.
Wilson, A. H., 1989. Distribution and controls of platinum group element mineralization in Cyclic Unit 1 of
the Great Dyke, Zimbabwe. In: Papunen, H. (ed.) Abstracts — 5th International Platinum Symposium. Bull.
Geol. Soc. Finland, 61 (Part 1), 3.
1990. Horizontal and vertical variations in noble metal distribution in the Great Dyke of Zimbabwe: a
model for the origin of the PGE mineralization by fractional segregation of sulfide. Chem. Geol. 88, 279-300.
Podmore, F., 1970. The shape of the Great Dyke of Rhodesia as revealed by gravity surveying. Spec. Publ. Geol. Soc.
S. Afr. 1, 610-20.
Wilson, A. H., 1987. A reappraisal of the structure, geology and emplacement of the Great Dyke, Zimbabwe.
In: Halls, H. C , & Fahrig, W. F. (eds.) Mafic Dyke Swarms. Geol. Assoc. Canada Spec. Paper 34, 317-30.
Prendergast, M. D., 1987. The chromite ore field of the Great Dyke, Zimbabwe. In: Stowe, C. W. (ed.) Evolution of
Chromium Ore Fields. New York: Van Nostrand Reinhold, 89-108.
1988. The geology and economic potential of the PGE rich Main Sulphide Zone of the Great Dyke,
Zimbabwe. In: Prichard, H. M., Potts, P. J., Bowles, J. F. W., & Cribb, S. J. (eds.) Geo-Platinum 87. London:
Elsevier, 281-302.
1989, The geology and stratigraphic setting of the Wedza-Mimosa platinum deposit, Great Dyke, Zimbabwe.
In: Papunen, H. (ed.) 5th International Platinum Symposium, Abstracts (poster). Bull. Geol. Soc. Finland 61,
13-14.
1990. Platinum group minerals and hydrosilicate 'alteration' in Wedza-Mimosa platinum deposit, Great
Dyke, Zimbabwe—genetic and metallurgical implications. Trans. Inst. Min. Metal. (Section B: Appl. Earth
Sci.) 99, 91-105.
1991. The Wedza-Mimosa platinum deposit, Great Dyke, Zimbabwe: layering and stratiform PGE
mineralization in a narrow mafic magma chamber. Geol. Mag. 128, 235-49.
Wilson, A. H., 1989. The Great Dyke of Zimbabwe. II Mineralization and mineral deposits. In: Prendergast,
M., & Jones, M. (eds.) 5th Magmatic Sulphides Field Conference, Harare, Zimbabwe. London: Institution of
Mining and Metallurgy, 21-42.
660 A L L A N H. W I L S O N

Robins, B., Haukvik, L., & Jansen, S., 1987. The organisation and internal structure of cyclic units in the
Honningsvag intrusive suite, north Norway: implications for intrusive mechanisms, double-diffusive convec-
tion and pore magma infiltration. In: Parsons, I. (ed.) Origin of Igneous Layering. NATO ASI Series. Dordrecht:
D. Reidel, 287-312.
Sparks, R. S. J., & Huppert, H. E., 1984. Density changes during fractional crystallization of basaltic magmas: fluid
dynamic implications. Contr. Miner. Petrol. 85, 300-9.
1987. Laboratory experiments with aqueous solutions modelling magma chamber processes. I. Discus-
sion of their validity and geological application. In: Parsons, I. (ed.) Origin of Igneous Layering. NATO ASI
Series. Dordrecht: D. Reidel, 527-38.
Kerr, R. C, McKenzie, D. P., & Tait, S. R., 1985. Postcumulus processes in layered intrusions. Geol. Mag.
122, 555-68.
Tait, S. R., Huppert, H. E., & Sparks, R. S. J., 1984. The role of compositional convection in the formation of
adcumulate rocks. Lithos 17, 139-46.
Turner, J. S., 1980. A fluid dynamic model of differentiation and layering in magma chambers. Nature 285, 213-15.
Campbell, I. H., 1986. Convection and mixing in magma chambers. Earth-Sci. Rev. 23, 255-352.

Downloaded from http://petrology.oxfordjournals.org/ at Aston University on January 30, 2014


Gustafson, L. B., 1978. The flow of hot saline solutions from vents in the sea floor—some implications for
exhalative massive sulfide and other ore deposits. Econ. Geol. 73, 1082-100.
1981. Fluid motions and compositional gradients produced by crystallization or melting at vertical
boundaries. J. Volcano!. Geothermal Res. 11, 93-125.
Ulmer, G. C , & Gould, D. P., 1982. Monomineralicity and oikocrysts: keys to cumulus cooling rates? Lunar Planet.
Inst. Tech. Rep. 80.01, 154.
Wager, L. R., & Brown, G. M , 1968. Layered Igneous Rocks. Edinburgh: Oliver & Boyd, 588 pp.
Wadsworth, W. J., 1960. Types of igneous cumulates. J. Petrology 1, 73-85.
Wagner, P. A., 1914. The geology of a portion of the Belingwe District of Southern Rhodesia. Trans. Proc. Geol. Soc.
S. Afr. 17, 39-54.
Weibe, R. A., & Wild, T., 1983. Fractional crystallization and magma mixing in the Tigalak layered intrusion, the
Nain anorthosite complex, Labrador. Contr. Miner. Petrol. 84, 327-44.
Weiss, O., 1940. Gravimetric and earth magnetic measurements of the Great Dyke of Southern Rhodesia. Trans.
Proc. Geol. Soc. S. Afr. 63, 143-51.
Whipple, E. R., 1974. A study of Wilson's determination of ferrous iron in silicates. Chem. Geol. 14, 223-9.
Wilson, A. H., 1977. The petrology and structure of the Hartley Complex of the Great Dyke, Rhodesia.
Unpublished D.Phil. Thesis, University of Rhodesia.
1982. The geology of the Great 'Dyke', Zimbabwe: the ultramafic rocks. J. Petrology 23, 240-92.
Naldrett, A. J., 1989. Vertical and lateral variations in the petrology, structure and mineral chemistry of Cyclic
Unit 1 of the Darwendale Subchamber of the Great Dyke, and their bearing on PGE and base metal
mineralization In: Papunen, H. (ed.) Abstracts—5th International Platinum Symposium. Bull. Geol. Soc.
Finland 61 (Part 1), 2-3.
Naldrett, A. J., & Tredoux, M., 1989. Distribution and controls of platinum group element and base element
mineralization in the Darwendale Subchamber of the Great Dyke, Zimbabwe. Geology 17, 649-52.
Prendergast, M. D., 1987. The Great Dyke of Zimbabwe — an overview. In: Campbell, A. (ed.) Field Guide for
the 5th Magmatic Sulphides Field Conference. IGCP project 181, 23-5.
1989. The Great Dyke of Zimbabwe. I: Tectonic setting, stratigraphy, petrology, structure, emplacement
and crystallization. In: Prendergast, M. D., & Jones, M. (eds.) 5th Magmatic Sulphides Field Conference,
Harare, Zimbabwe. London: Institution of Mining and Metallurgy, 1-20.
Tredoux, M., 1990. Lateral and vertical distribution of the platinum-group elements and petrogenetic controls
on the sulfide mineralization in the PI pyroxenite layer of the Darwendale Subchamber of the Great Dyke,
Zimbabwe. Econ. Geol. 85, 556-84.
Wilson, J. F., 1981. The Great 'Dyke'. In: Hunter, D. R. (ed.) Precambrian of the Southern Hemisphere.
Amsterdam: Elsevier, 572-8.
Wilson, J. R., & Larsen, S. B., 1985. Two-dimensional study of a layered intrusion in the Hyllingen Series, Norway.
Geol. Mag. 122, 97-124.
Menuge, J. F., Pedersen, S., & Engel-Sorensen, O., 1987. The southern part of the Fongen-Hyllingen layered
mafic complex, Norway: emplacement and crystallization of compositionally stratified magma. In: Parsons, I.
(ed.) Origin of Igneous Layering. NATO ASI Series. Dordrecht: D. Reidel, 145-84.
Worst, B. G., 1958. The differentiation and structure of the Great Dyke of Southern Rhodesia. Trans. Geol. Soc. S.
Afr. 61, 283-354.
1960. The Great Dyke of Southern Rhodesia. Bull Geol. Surv. S. Rhodesia 47, 234 pp.

APPENDIX
Calculation of residual porosity and cumulus components
The method of calculation is based on achieving mass balance between the intercumulus liquid and
closed-system crystallization of orthopyroxene, clinopyroxene, and plagioclase, and broadly follows
THE GEOLOGY OF THE GREAT DYKE, ZIMBABWE 661

the approach of Barnes (1986). In contrast to the approach of Barnes (1986), no prior assumption is
made regarding the porosity at the instant the cumulus system becomes closed (residual porosity). The
composition of the original liquidus orthopyroxene (which is determined by the intercept on the mg-
number vs. P plot) becomes modified as a result of equilibrium or fractional crystallization of the
trapped liquid to form a zoned overgrowth on the cumulus orthopyroxene, followed by solid-state
homogenization to remove the resultant compositional gradients. As cooling is relatively slow, re-
equilibration both within and between crystals, and with the liquid will take place before solidification
of the trapped liquid is complete. The overgrowth of the cumulus phases may therefore be regarded as a
combination of both fractional and equilibrium crystallization, and the final homogeneous composi-
tion will be the same irrespective of which control was dominant and the relative influences of the two
processes.
In this model an appropriate starting liquid [such as that modelled by Wilson (1982) for the base of
the PI layer] is adjusted by fractional crystallization of orthopyroxene until it is in equilibrium
with the liquidus composition of the orthopyroxene. It is also assumed that clinopyroxene joins

Downloaded from http://petrology.oxfordjournals.org/ at Aston University on January 30, 2014


orthopyroxene on the liquidus soon after solidification of the trapped liquid commences and that the
ratio of orthopyroxene to clinopyroxene is 1:1. The form of the zoned compositional profile around
the unzoned liquidus core is essentially constant for the given mafic assemblage and only the width of
the zone varies. The width of the zoned margin is adjusted (in terms of fraction of the total mass of the
final orthopyroxene) until the bulk composition of the orthopyroxene is the same as that measured in
the rock. The proportion of orthopyroxene that has been added to the original cumulus crystal by
crystallization of the intercumulus liquid can therefore be calculated.
The mass balance requirement of the model is such that all the trapped liquid must have crystallized
at the same instant that all the MgO is used up in the crystallization of pyroxenes. The ratio of the
mafic phases to plagioclase crystallizing from the liquid is adjusted until this condition is met.
Distribution coefficients for Mg and Fe 2 + between ortho- and clinopyroxene and liquid are those
determined by Wilson (1982) for the Great Dyke for coexisting orthopyroxene and olivine, and for
coexisting ortho- and clinopyroxene. The calculated proportion of plagioclase to pyroxene forming
from the trapped liquid is 30-50%. The normative composition of the final amount of trapped liquid
in most cases is in close agreement with the observed late-stage assemblage of quartz, K-feldspar,
albitic plagioclase, magnetite, and phlogopite.
The difference between the observed amounts of discrete postcumulus phases in the rock and the
calculated residual porosity (minus the orthocumulus component of the orthopyroxene) represents the
amount of oikocrystic phases which developed in equilibrium with the main body of magma (i.e., as
heteradcumulus components of plagioclase and clinopyroxene). These phases would also have been
zoned during the solidification of the trapped liquid. The entire mineralogical budget of the
orthopyroxene cumulate can be expressed as
cumulus Opx + heteradcumulus Cpx and/or Plag
(adcumulus components in equilibrium with bulk liquid)
+
overgrowth on cumulus Opx and heteradcumulus phases + new interstitial phases
(orthocumulus components resulting from solidification of the trapped liquid)

In addition, the amount of discrete postcumulus phases can be expressed as


heteradcumulus phases
+
overgrowth on heteradcumulus phases
+
new interstitial phases.
The calculated residual porosity and the amount of heteradcumulus components, as well as the
calculated proportion of the crystallizing mineral phases, is dependent on the total amount of MgO in
the liquid, but not markedly so. However, these parameters are strongly dependent on the original and
final compositions of orthopyroxene and the final modal percentage (converted to weight per cent) of
cumulus orthopyroxene in the rock, which are uniquely determined for each sample, and therefore
errors introduced from these parameters will be relatively small.
Table A1 shows the effect on the modelled parameters of errors in the estimated MgO content of the
primary liquid. Errors introduced in the calculated residual porosity and heteradcumulus component
662 ALLAN H. WILSON

are, in general, small within the limits of petrogenetic modelling, and in any case, most applications will
be for comparative purposes within a suite of samples for a given layer.
Calculated residual porosity and measured modal percentage of orthopyroxene are related (Fig. Al)
in terms of the final re-equilibrated composition of orthopyroxene. Mass balance is required between

TABLE Al
Influence of initial liquid composition on modelled parameters

Initial liquid Orthopyroxene Modelled parameters


% FeO % MgO MgO/FeO mg-number rag-number Modal % % Trapped % Heteradcumulus
(liquidus) (final) liquid component

6 61 11-93 1-80 0-865 0-860 50 1-9 48-7

Downloaded from http://petrology.oxfordjournals.org/ at Aston University on January 30, 2014


8-29 9-79 118 0-865 0-860 50 1-4 49-2
6-61 11-93 1-80 0-865 0-840 80 15-2 9-4
8-29 9-79 118 0-865 0-840 80 11-3 13-3

= 0.045
Mg# = 0.82

= 0.035
Mg# = 0.83

= 0.025
Mg# = 0.84

= 0.015
Mg# = 0.85

= 0.005
Mg# = 0.86
AMg# = O
Mg#° = 0.865
40 50 60 70 80 100
% Orthopyroxene (mass) observed
in sample
FIG. Al. Calculated residual porosity and measured mass percentage of cumulus orthopyroxene related to
compositional changes by reaction with trapped liquid. The initial liquidus composition of pyroxene is kept
constant [mj-number 0 = 0-865] and the re-equilibrated composition (mg-number) and change (Amg-number) are
shown for changing residual porosity. In the extreme case of zero porosity, required for a perfect adcumulate (line
A-B), the original liquidus composition is unchanged and all discrete postcumulus phases must be heteradcumulus.
For all other situations, mass balance is required between the final amount of orthopyroxene, the change in
composition, and the heteradcumulus component of the discrete postcumulus phases. The example given by point
C is for a cumulate which has 8 1 % orthopyroxene of liquidus composition mg-number 0-865, and a residual
porosity of 16%. The re-equilibrated composition is mg-number 0-83, and 10% of the rock comprises heteradcumu-
lus component as part of the discrete postcumulus assemblage.
THE GEOLOGY OF THE GREAT DYKE, ZIMBABWE 663

the amounts of orthopyroxene (before and after postcumulus overgrowth), residual porosity, and the
heteradcumulus component (oikocryst) of the discrete postcumulus assemblage. As the percentage of
orthopyroxene in the rock decreases, then the lower porosity is more effective in modifying the
composition of the orthopyroxene because of mass balance requirements. For a residual porosity of
10% a change in mg-number of 0045 could occur for a rock with 40% orthopyroxene as compared
with a change in mg-number of 0-02 for a rock which has 85% orthopyroxene. The magnitudes of the
compositional shifts are similar to those suggested by Barnes (1986).

Downloaded from http://petrology.oxfordjournals.org/ at Aston University on January 30, 2014

You might also like